탐색으로 건너뛰기

Scribe v3

Scribe v3 and Scribe v3 Live, the most accurate ElevenLabs speech recognition models, are now available in the Speech to Text API. Scribe v3 (scribe_v3) transcribes recorded audio, and Scribe v3 Live (scribe_v3_live) transcribes in real time. They succeed Scribe v2 and Scribe v2 Realtime, which remain available.

Scribe v3 makes 27% fewer word errors than Scribe v2 on the FLEURS benchmark across 102 languages and labels speakers more accurately. It adds numbers_format to write numbers as digits or words, and secondary_language_codes for audio that mixes languages. Scribe v3 Live has a 44% lower semantic word error rate than Scribe v2 Realtime on FLEURS across 73 languages. It adds turn prediction, which commits a segment when the speaker has finished their turn instead of at the first pause, a multilingual mode that writes each language in its own script, and a keyterm library of up to 10,000 terms.

Unity SDK

The ElevenAgents Unity SDK is now available as the io.elevenlabs.agents package, starting with v0.1.0. It supports the Unity Editor, desktop standalone players on macOS, Windows and Linux, and WebGL builds, with iOS, Android and XR in beta. It requires Unity 6.3 LTS or later.

The SDK connects a Unity scene to an agent through a single Conversation API. It captures the player’s microphone, plays the agent’s voice through an AudioSource you can position in the scene for spatial audio, and raises transcript, response, mode and tool call events on Unity’s main thread. Desktop and mobile builds connect over WebSocket, and WebGL builds can also use WebRTC.

Workflows as code

ElevenAgents workflows can now be built in code with the Workflow builder in the Python and JavaScript SDKs, from v2.77.0. Chain converse, tool, transfer and end steps with when, on_success and go_to routing, then pass the result as workflow to agents.create() or agents.update(). The builder runs entirely in the SDK, so you can load an agent’s workflow, change one node, keep workflows as JSON in version control and still edit them in the visual editor. See Workflows as code.

More agent tools in the hosted MCP server

Hosted MCP adds 24 tools for inspecting agents, covering analytics, configuration summaries, insights, alerts, test summaries, branch conflicts and conversation counts. They use the existing convai_read and convai_write permissions, so connected MCP clients can review an agent the way Architect does. MCP clients can also find and replace or append text in an agent’s system prompt or first message, staged in a branch draft until it is published. See Hosted MCP.

All changes

  • Workflows as code: The Python and JavaScript SDKs add a Workflow builder for agent workflows. See Workflows as code. The workflow editor adds a minimap, zooms out further and has a snap-to-grid toggle, and exposes the subagent turn.soft_timeout_config override so dashboard saves no longer remove it.
  • Clone strength and reasoning effort overrides: Agent tts.clone_strength, clone_strength on supported voices and a clone_strength conversation override are now in the public API. Clients can set conversation_config_override.agent.prompt.reasoning_effort once the agent owner enables the Reasoning effort override; a level the agent’s LLM does not support stops the conversation from starting. Realtime agents show the stability, similarity and clone strength sliders. See Overrides.
  • Conversation data: List conversations returns cost (credits) and cost_fiat (USD), and the conversations table has an optional Cost column. Transcript rows on asynchronous channel conversations add created_at_unix_ms, and asynchronous conversation metadata adds integration_trigger_id and platform_metadata, such as the Zendesk ticket_id. Post-call audio and unredacted audio webhooks include branch_id, version_id and environment.
  • Conversation history limit: Conversation history, count and search requests from users who can see more than 10,000 agents return 403 conversation_history_agent_limit_exceeded unless they filter by agent, instead of returning truncated results.
  • Guardrails: Triggered guardrails record execution_mode (blocking or streaming) and guardrail_details, with the judge’s rationale or the classifier verdict, and the conversation transcript shows why a guardrail triggered. Test runs return triggered_guardrails and show guardrail triggers in the test transcript. See Guardrails.
  • Branches: List branches and the branch get and update endpoints return sync_status for branches that update automatically, with the outcome and attempted_at time of the last automatic rebase, and the branches table shows its state. Open merge proposals whose target branch is merged and archived are retargeted to the merge target instead of closing. Shareable agent links follow the live traffic split across branches by default, with pinning to the current branch as a separate option.
  • Defaults and models: Agents created with Create agent have workflow.prevent_subagent_loops on unless the request sets it. When Eleven v3 Conversational fails during a call, agents fall back to Eleven v4 Turbo, Eleven v4 and then Flash, and never to eleven_v3, so the voice no longer changes mid-call. The fallback ramp for gemini-2.5-flash and gemini-2.5-flash-lite now starts on November 15 instead of November 8; the November 30 cutoff is unchanged.
  • Realtime API: response.created and response.done always include response.status_details. turn_detection.interrupt_response: false is honored, and only committed agent text is streamed. audio/pcm input at sample rates other than 24000 Hz is rejected with invalid_audio_format, and audio/pcma is removed from the audio format enum. A function_call_output that arrives after its call stopped waiting no longer suppresses the next response.create. See Realtime models.
  • Knowledge base: Sitemap crawls have an Auto Discover option that adds new sitemap pages at the folder’s auto-sync frequency. Test retrieval in the dashboard covers every retrieval type. The knowledge base tool is registered only when the agent has documents attached. See Knowledge base.
  • Channels and telephony: Pre-call webhook request_headers accept environment variables, in the API and the dashboard. The Salesforce Cases trigger accepts an optional Reply-To address for email replies. Activating a second Telegram trigger for the same bot returns 409 instead of replacing the first. Redelivered WhatsApp messages no longer get a second reply. Amazon Connect calls no longer end with 408 when the agent pauses. transfer_to_agent conditions resolve dynamic variables. WebRTC calls drop stale audio after an interruption, and an inbound audio batch longer than 15 minutes keeps the newest 15 minutes instead of ending the call.
  • Conversation WebSocket: Text frames that are not a JSON object close the connection with 1008 instead of 1011. Error closes give end users a generic reason; the detailed reason is sent only to sessions of the agent’s owner.
  • Privacy: Conversation history redaction runs on each turn of asynchronous channels such as Slack, email, WhatsApp text and Telegram before the turn is stored. With history redaction on and audio recording off, audio post-call webhooks and storage exports deliver entity-redacted audio. Triage no longer clusters issues from agents in zero retention mode workspaces.
  • Dashboard: Agents Home (/app/agents) opens Architect, and the previous Home dashboard moves to Dashboards (/app/agents/analytics). Architect’s Pro and Lite menu lets you choose a reasoning effort on every plan. Concurrency charts count only voice conversations and no longer change with the chart granularity. Saved views keep every filter, agent search matches partial words in agent names, the Tools tab groups statistics by tool ID, and batch call CSV exports work while calls are still in progress. Triage tickets can be sorted by last_reported. show_rich_content carousels accept up to 10 cards, up from 5.
  • Evaluations and analysis: Run conversation evaluation accepts evaluated_agent_id and evaluated_branch_id, so reruns work for conversations transferred to another agent; rerunning requires edit permission. Analysis item references without a pinned version store version_id as "latest". Tool latency statistics used for auto pre-tool speech exclude MCP approval waits and failed calls.
  • Fixes: Conversations on Claude 5.5 models no longer fail with 400 when a response interleaves thinking with text or tool calls. Test runs from the agent editor compile procedures, so structured procedures no longer fail with a transfer error. Merges and rebases no longer report conflicts when both branches made the same procedure change, or on computed test fields and MCP settings. Duplicating an agent remaps procedure references in structured procedure steps. Conference transfers on outbound SIP trunk calls work when the number has no assigned agent. Workflow nodes with tool overrides that previously failed to save now save, new backward transitions no longer show an exit_slot error, and enum values on LLM expression schemas are kept on save. Calls whose environment does not exist are marked failed with the reason in metadata.error. Architect works again in the India and Singapore data residency environments.
  • Credit limits: PUT /v1/workspace/members/{user_id}/usage-limit sets the monthly credit limit (usage_limit) of a member or service account, or removes it with null. Admins can also set it from the Credit limit button on the service accounts page. Personal credit limits now count API key and agent usage for every seat type except basic, which still counts only Creative usage. API key and agent usage from before this change is not counted retroactively. See Members.
  • Billing group concurrency: Enterprise admins can set Text to Speech, Dubbing and Music concurrency limits on billing groups from the group dialogs. See Billing groups.
  • Architect permission: Existing full and read-only Agents grants, including grants to Everyone and to groups, now also grant the Architect permission.
  • Auth connections: OAuth2 JWT auth connections add jwt_usage, where client_authentication sends the JWT as a client_assertion on a client_credentials request. OAuth2 JWT and private key JWT connections add extra_headers for additional JWT header parameters, such as x5t, and the PS256, PS384 and PS512 algorithms. See Create auth connection.
  • Webhooks: OAuth2 webhooks can send client credentials in a Basic authorization header (token_endpoint_auth_method: client_secret_basic). GET /v1/workspace/webhooks?include_usages=true no longer times out for workspaces with many agents, and webhook retries no longer deliver the same job twice.
  • Audit logs: The activity_name and class_name query parameters of List audit logs are typed as the AuditActivityName and AuditClassName enums, which list every accepted value.
  • Fixes: SCIM requests that time out waiting on a group return 429 with Retry-After: 5 instead of 500. Saving the Everyone group no longer fails in workspaces with a permission allowlist.
  • Concurrency limits: Sound effects have their own per-plan concurrency limits instead of the Text to Speech limit: 4 on Free and Starter, 6 on Creator and Pro, and 8 on Scale and Business. Requests beyond the limit and its reduced-priority queue return 429. Enterprise customers can request a higher limit. The usage analytics concurrency chart shows Sound effects with its limit. See Sound effects.
  • Authentication: auth.enable_auth (boolean, default false) on Create Speech Engine and Update Speech Engine requires a signed URL or token to start a conversation. The Security tab has an Enable authentication toggle.
  • Sharing: Workspace sharing endpoints accept a Speech Engine’s seng_ ID, so Speech Engines can be shared in a workspace.
  • Agent inspection tools: 24 tools for agent analytics, configuration summaries, insights, alerts, test summaries, branch conflicts and conversation counts, available with the existing convai_read and convai_write permissions. See Hosted MCP.
  • Prompt editing: MCP clients can find and replace or append text in an agent’s system prompt or first message on a branch. The edit is staged in the branch draft and goes live when the draft is published.
  • Paginated tool listing: tools/list is returned in pages of at most 256 KiB, so MCP clients that reject large listings can connect.
  • Fixes: OAuth clients that open the authorization page in a popup, such as Gemini Enterprise, can connect. Media in the Flows view loads in ChatGPT.
  • Scribe v3 and Scribe v3 Live: Create transcript accepts scribe_v3 as model_id, with new numbers_format (digits or words) and secondary_language_codes parameters. The realtime WebSocket accepts scribe_v3_live, with commit_strategy=turn_prediction, multilang_mode and a keyterm library of up to 10,000 terms. Scribe v2 and Scribe v2 Realtime remain available. See Models.
  • Timestamps with background filtering: The realtime WebSocket accepts include_timestamps together with filter_background_audio, with timestamps mapped to the original audio.
  • Lower partial latency: Realtime partial transcripts are sent as soon as they are produced, removing up to 100 ms of delay. turn_probability on partial transcripts is rounded to two decimal places.
  • EU data residency: transcript_edit is available in the EU data residency environment.
  • Fixes: Realtime transcripts no longer contain emotion labels such as {neutral}.
  • Voice Library matches: When Instant Voice Clone samples match a protected voice that is available in the Voice Library, Create IVC voice and POST /v2/voices/create-preview-from-samples return library_voice_match, with library_voice_id and public_owner_id. If the clone is blocked, the 403 response includes the same object in detail.data. Pass both IDs to POST /v1/voices/add/{public_user_id}/{voice_id} to add the library voice instead.
  • Voice categories: The VOICE_CATEGORY enum no longer lists famous.
  • Fixes: Instant Voice Clone drafts are stored per user and workspace, and the page no longer loops on an old draft. Adding a library voice to My Voices no longer reports success without listing the voice.
  • Failed languages: Target languages that cannot start, for example because the quota is exceeded, are marked FAILED instead of staying PENDING, and targets pending for more than 7 days fail.
  • Short sources: Sources shorter than 0.1 seconds are rejected with source_too_short before any charge.
  • Verified clones: Dubbing accepts a workspace’s own verified voice clones that match a protected voice instead of blocking them.
  • Legacy manual mode: The Manual mode option is renamed Legacy manual mode and is shown only when Dubbing v1 is selected.
  • Aspect ratios: Projects can use 3:4 and 4:3 aspect ratios, and media in those ratios starts new projects at the matching ratio.
  • Subtitle timing: SRT and VTT exports apply paragraph trims, so subtitles match the exported audio.
  • Faster blank projects: Blank audiobook projects are created without a wait, and blank eleven_v3 projects no longer get the sample template.
  • Fixes: Replacing the content of a large project with POST /v1/studio/projects/{project_id}/content no longer fails with 500. IME input, such as Japanese, is no longer committed early or mid-word in the editor. Pause after paragraph pauses at the real end of a paragraph that is still loading, and no longer pauses as soon as you press Play at the start of a paragraph.
  • Image editor: Rotating a layer rotates its selection box, masks sit in their own draggable box, and text and shape layers can be rotated, masked and recolored.
  • Flows: Figma .fig imports keep small icons whole and keep drop shadows, rounded clipping and borders of stroke-only frames. The Text to Speech node offers the full 0 to 100% Stability range for Eleven v4. Seedance 2.5 unlocks the aspect ratio when a video reference is selected, and dropped assets are arranged in a grid.
  • Video in Voice Changer: Voice Changer in the app keeps video uploads as video. History opens a viewer with the original and converted audio tracks, and the converted video can be downloaded as MP4.
  • Fixes: Voice Isolator returns 400 invalid_audio instead of 500 for pcm_s16le_16 uploads with an odd number of bytes.
  • Pronunciation previews: The pronunciation dictionaries dialog has a model picker for previewing rules, instead of always previewing with eleven_turbo_v2.
  • Voice library surcharge: The agents voice library surcharge is charged per voice, so usage analytics attributes it to each voice instead of “No voice”.
  • AI Speech Classifier: The classifier accepts video files and checks their audio track.

JavaScript SDK

  • v2.77.0 - Added a Workflow builder, exported from the package root, that builds agent workflows with chained calls and produces the workflow value for agents.create() and agents.update(). See Workflows as code.

Python SDK

  • v2.77.0 - Added the Workflow builder in elevenlabs.agent_workflows, which builds agent workflows with chained calls and produces the workflow value for agents.create() and agents.update(). Speech Engine create and update accept the auth parameter.

ElevenLabs CLI

  • v1.9.1 - agents workflow diagram draws bare workflow files exported by the SDK Workflow builders. onboard waits up to 30 minutes for sign-in approval instead of 5, retries a token request that could not connect, and reports a copied key that /v1/user refuses as “couldn’t be checked” instead of belonging to another account.

Unity SDK

  • v0.1.0 - First release of the ElevenAgents Unity SDK (io.elevenlabs.agents), installed from the git URL through the Unity Package Manager. Adds the Conversation session API, conversation events, messaging, client tools, audio and device controls, a native WebSocket transport for the Editor and desktop players, and a WebGL transport with WebSocket and WebRTC. Requires Unity 6.3 LTS or later. See Unity SDK.

Packages

New endpoints

Workspaces

  • PUT /v1/workspace/members/{user_id}/usage-limit - Set or clear the monthly credit limit of a workspace member or service account. Requires usage_limit (integer, minimum 0, or null to remove the limit). Admins only.

Updated endpoints

ElevenAgents

  • Create agent and the agent get and update endpoints
    • TTS configuration adds clone_strength (number, default 0), supported voices add nullable clone_strength (number), and TTS conversation and workflow overrides add nullable clone_strength. The override config adds tts.clone_strength (boolean, default false).
    • Prompt overrides add nullable reasoning_effort (LLMReasoningEffort), and the override config adds prompt.reasoning_effort (boolean, default false). conversation_config_override on the /v1/convai/conversation WebSocket accepts agent.prompt.reasoning_effort and tts.clone_strength.
    • Pre-call webhook request_headers values accept ConvAIEnvVarLocator in addition to strings and secret locators.
    • Analysis item references (AttachedUserEvaluationRef, AttachedUserDataCollectionRef) default version_id to "latest" and it is no longer nullable.
    • SearchStrategy defaults to semantic.
    • DataCollectionValueSchema.allowed_values_dynamic_variable is deprecated; use allowed_values instead.
  • List conversations - Added nullable cost (integer) and cost_fiat (number) to each conversation summary.
  • Get conversation details
    • Asynchronous conversation metadata adds nullable integration_trigger_id (string) and platform_metadata.
    • Transcript rows add nullable created_at_unix_ms (integer), also in test responses.
    • Triggered guardrails add nullable execution_mode (streaming or blocking) and guardrail_details (LLM, classifier or unknown verdict).
    • AgentDefinitionSource adds mcp.
  • Run conversation evaluation - Added optional evaluated_agent_id and evaluated_branch_id (string).
  • List branches and the branch get and update endpoints - Added nullable sync_status with last_attempt (outcome: succeeded, conflict, no_common_ancestor, rejected or error, and attempted_at in Unix seconds).
  • Test runs (POST /v1/convai/agents/{agent_id}/run-tests and the test invocation endpoints) - Added triggered_guardrails (array).
  • Triage tickets (GET /v1/convai/triage-tickets and GET /v1/convai/agents/{agent_id}/triage-tickets) - The sort enum adds last_reported.
  • Realtime client secrets (POST /v1/convai/realtime/client_secrets) - AudioFormat no longer lists audio/pcma.

Speech Engine

Voices

  • Create IVC voice - Added nullable library_voice_match with library_voice_id and public_owner_id (string).
  • VOICE_CATEGORY, used by the dubbing similar voices endpoint, removed famous.

Workspaces

  • Create auth connection and the auth connection get and update endpoints
    • OAuth2 JWT connections add jwt_usage (authorization_grant, the default, or client_authentication).
    • OAuth2 JWT and private key JWT connections add extra_headers (map of strings), and algorithm adds PS256, PS384 and PS512.
  • List audit logs - activity_name and class_name query parameters changed from string to the AuditActivityName and AuditClassName enums.

SFX v3

SFX v3 (eleven_text_to_sound_v3), our newest sound effects model, is now available to every workspace. Compared with eleven_text_to_sound_v2, it follows prompts that contain several instructions more closely, handles prompts with music or vocals better, sounds more natural and is less likely to clip on high-frequency sounds such as chimes and whistles.

In ElevenCreative, SFX v3 is the default model in Sound Effects, Studio and Flows, and the sound effects that Video to music adds are generated with it. In the API, set model_id to eleven_text_to_sound_v3 on Create sound effect, and set loop to true for a sound effect that loops smoothly. eleven_text_to_sound_v2 is deprecated and will be removed on March 1, 2027.

Eleven Realtime for every workspace

ElevenAgents voice agents in every workspace can now use Eleven Realtime, completing the gradual rollout that started on October 19. Pick Eleven Realtime v1 Mini or v1 Max as the agent’s voice mode, or connect an OpenAI Realtime client to the Realtime WebSocket, and review sessions on the Realtime Conversations page. Realtime conversations can also be organized into groups through the API, under /v1/convai/realtime/conversation-groups.

Professional Voice Clones on Eleven v4

Text to Speech Professional Voice Clone fine-tuning for Eleven v4 and Eleven v4 Turbo is now available to every workspace. To train an existing clone, hover over it in My Voices and click the plus button next to Eleven v4. Requests for a clone fine-tuned on v4 now use the clone_strength saved in the voice’s settings when the request sets neither clone_strength nor use_pvc_as_ivc, so you no longer need to pass it on every call. See Eleven v4.

Analysis items API

ElevenAgents evaluation criteria and data collection items can now be managed through the API as reusable analysis items. Create, list, update, copy and delete them under /v1/convai/analysis-items, attach or detach them across agents and branches with a single request, list each item’s versions, and check which agents depend on an item before you change or delete it. Inline evaluation criteria and data collection in agent configurations keep working and are stored as attached items. See Analysis library.

Noisy environment detection in the API

ElevenAgents noisy environment detection can now be configured through the API and CLI. Set conversation_config.turn.noisy_environment_detection to true and the agent speaks noisy_environment_message once, without interruption, when it hears other people speaking in the caller’s background, asking them to move somewhere quieter. Add a translated message for each additional language in language_presets. The setting is available for voice agents in Enterprise workspaces; see Noisy environment detection.

Realtime entity redaction

Speech to Text can now redact entities in realtime transcripts. Pass entity_redaction and entity_redaction_mode as query parameters on the realtime WebSocket, in the same format as batch transcription, and each committed transcript is followed by a redacted_transcript event carrying the original text and the redacted_text. The Python and JavaScript SDKs from v2.76.0 and @elevenlabs/client from 1.34.0 support these options. See Entity detection.

All changes

  • Analysis items API: Evaluation criteria and data collection items can be managed as reusable analysis items under /v1/convai/analysis-items, with endpoints to list, create, get, update, copy and delete items, list versions, list dependent agents and update attachments in bulk.
  • Noisy environment detection: turn.noisy_environment_detection (default false) and turn.noisy_environment_message are available in the API, along with noisy_environment_message_translation on language presets and noisy_environment_message in conversation and workflow overrides. Available for voice agents in Enterprise workspaces. See Noisy environment detection.
  • Eleven Realtime available to all workspaces: Realtime voice mode, the Realtime WebSocket and the Realtime Conversations page are available to every workspace, completing the gradual rollout announced on October 19. Realtime conversation groups can now be managed through the API, under /v1/convai/realtime/conversation-groups.
  • Realtime API behavior: Each manual input_audio_buffer.commit transcribes only the speech since the previous commit, and input_audio_buffer.clear drops buffered audio. response.metadata from response.create is echoed on response.created and response.done, with at most 16 string pairs. input_audio_buffer.speech_started and speech_stopped are sent once per committed user turn instead of on every pause. Idle audio sessions close after 3 minutes instead of 60 seconds, after a session_idle_timeout error.
  • GPT-6.1 Terra: gpt-6.1-terra can be selected as the agent LLM and as the analysis LLM, on the standard and fast service tiers, with reasoning effort from none to max (default medium).
  • Default LLM: New agents created without an explicit LLM use gemini-3.5-flash instead of gemini-2.5-flash, and the default backup LLM cascade leads with Gemini 3.5 Flash models. Existing agents are unchanged. Gemini 3.5 Flash-Lite is also available in the India and Singapore isolated environments.
  • Model deprecations: gpt-4.1-nano and gpt-4o-2024-05-13, which OpenAI shut down on October 23, are deprecated immediately, and agents using them run on gpt-5.6-luna and gpt-5.6-sol respectively. gpt-5, gpt-5-mini, gpt-5-nano and their 2025-08-07 snapshots are deprecated ahead of OpenAI’s December 11 shutdown, with traffic moving to gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna in steps from 10% on November 11 to 100% on December 6. List LLMs returns deprecation_info for these models. Claude Sonnet 4.5 is deprecated on November 30, with Claude Sonnet 4.6 as the replacement, and is removed from model selectors. The fallback ramp for gemini-2.5-flash and gemini-2.5-flash-lite now starts on November 8 instead of November 1; the November 30 cutoff is unchanged.
  • Conversation message counts: List conversations and Get conversation details return user_message_count and agent_message_count. message_count on the conversation list now counts only turns with text, so tool-only turns are no longer counted. Conversations with no messages get a call_successful value of unknown, and the dashboard shows them as “No conversation”.
  • Knowledge base: Update document accepts enable_auto_sync, auto_remove and minimum_frequency_days, and the dashboard can edit these auto-sync settings after a document is created, optionally applying folder settings to existing documents. Sitemap crawls accept exclude_patterns, with an Exclude patterns field in the sitemap step of the URL dialog. Website crawls deduplicate pages by canonical URL, and Google Drive auto-sync pauses when the Drive token is rejected and resumes after you reconnect. See Knowledge base.
  • Agent testing: Tool call tests accept guardrail_trigger to assert that a guardrail should or should not trigger on the generated turn, with Should trigger and Should not trigger options in the dashboard. Tests whose turn is blocked by a guardrail now fail with “Agent response blocked by guardrail” instead of returning an inconclusive result. List test invocations accepts created_by_user_id, and the test history adds a creator filter. Test runs are returned in creation order. See Agent testing.
  • Branches and merges: Branches can have a default environment, set when creating or publishing to a branch and changed from the branch menu. Conversations on the branch use that environment unless the caller requests another one, and such branches cannot receive live traffic. Merge and rebase conflicts can be resolved with Architect, and procedure conflicts can be edited and combined like prompt conflicts. A single procedure’s unpublished changes can be discarded without discarding the whole draft. Branches whose only version was created with changes can be merged, and branch responses report a current_live_percentage of 100 for the main branch when the agent has no live deployment.
  • Workflows: Update state nodes and the update_state tool can override LLM settings (model, temperature, token limit and reasoning) for the rest of the conversation from a new LLM tab. Double-click a workflow edge to add reroute points that shape it. Workflow tool nodes accept only configurable system tools plus start_procedure and end_procedure.
  • Issues: Issue descriptions can be edited from the issue details sheet, and Update triage ticket accepts qa_comment. See Triage.
  • WhatsApp: WhatsApp conversations record call_result (answered, voicemail, busy, no_answer or failed), and conversations opened with a template record opening_message with the template name, delivery timestamps, error code and pricing category. Call permission requests record responded_at_unix, permanent and granted_by_user_call. WhatsApp batch calls that are declined, missed or cannot be placed mark the recipient failed instead of leaving it initiated. See WhatsApp.
  • Widget: Widget configuration adds transcript_speech_sync_enabled (default true), with a Reveal transcript as the agent speaks switch; set it to false to show each message in full at once. show_rich_content carousel cards accept rating (0 to 5) and review_count, shown as stars. Resumed persistent text conversations replay rich content in their history. See Widget.
  • Channel integrations: Asynchronous channels such as Slack, Telegram, Intercom, Gmail, Salesforce and Custom Channel record message timestamps from the source platform and store the conversation’s main_language from the language override or the agent’s language instead of always en. Salesforce passes the customer’s email to the agent as integration__salesforce_customer_email, which is not a verified identity. Gmail triggers run on the agent’s current version, back off on Gmail rate limits and set aside threads that keep failing.
  • Telephony: Twilio conference transfers can combine post-dial digits with call screening (require_acceptance). Tools that wait for speech to finish, such as transfer_to_number and end_call, are skipped if the caller hangs up during their pre-tool speech. Calling skip_turn after a turn timeout starts the skip-turn wait, so the silence timeout no longer ends the call during it.
  • Environment variables: Deleting an environment variable that the tip of a live branch references fails with 409 and lists the affected agents and branches. The environment variables table adds Remove value and Delete row actions.
  • Dashboard: The conversation view shows human agent replies captured after an escalation in a separate post-escalation section. The system prompt editor’s wand opens an Ask Architect menu to review and improve the prompt or write a new one. Filtered conversation exports include evaluation criteria and data collection results. Turns where the agent speaks after a silence timeout no longer report end-to-end latency. The agent voice settings show the Clone strength slider for every voice fine-tuned for a v4 model.
  • Fixes: Files that the LLM provider cannot process are dropped instead of failing later turns. Deleting an agent removes its trigger connections, so the integration connection can be deleted afterwards. The prompt editor saves & and > as typed instead of encoding them. Say nodes no longer duplicate messages, merge proposals that add a node group no longer fail, and node groups show their configured titles. The structured procedure editor opens procedures authored through the API, and dynamic variables in If step expressions are included in referenced_dynamic_variables. Agents whose data collection item uses a dynamic variable can be published again. Knowledge base rediscovery adds pages to older crawl jobs. The concurrency chart no longer overstates peaks when grouped by agent or branch. Garbled characters in LLM tool call arguments are repaired, and WhatsApp webhooks with resent interactive replies or contacts without a name are accepted instead of failing with 422.
  • Realtime entity redaction: The realtime WebSocket accepts entity_redaction (all, a category such as pii, phi or pci, or a list of entity types) and entity_redaction_mode (redacted, entity_type or enumerated_entity_type, the default). Each committed transcript is followed by a redacted_transcript event, and the settings are echoed in session_started. entity_redaction cannot be combined with transcript_edit. See Entity detection.
  • Realtime authentication: Bearer tokens passed to the realtime WebSocket in Sec-WebSocket-Protocol require the Speech to Text permission, as the other authentication methods do.
  • Usage on empty responses: Empty and silent responses from Create transcript include audio_duration_secs.
  • Fixes: Webhook transcription requests, with a file upload or cloud_storage_url, now apply diarization_threshold and seed.
  • SFX v3 for everyone: SFX v3 (eleven_text_to_sound_v3) is available to every workspace and is the default model in Sound Effects, Studio and Flows. The sound effects that Video to music adds with add_sfx are generated with SFX v3. See SFX v3.
  • Professional Voice Clones on Eleven v4: Professional Voice Clone fine-tuning for Eleven v4 and Eleven v4 Turbo is available to every workspace. See Eleven v4.
  • Saved clone strength: Eleven v4 requests for a fine-tuned Professional Voice Clone use the voice’s saved clone_strength when the request sets neither clone_strength nor use_pvc_as_ivc.
  • Pronunciation dictionary limit: Applying pronunciation dictionaries can expand the input text to at most 50 times its length plus 10,000 characters. Requests over the limit return 400 pronunciation_dictionary_expansion_too_large.
  • Fixes: History items for v4 generations keep clone_strength and speed in their saved settings.
  • Eleven v4 recommended: Project settings and the model picker recommend Eleven v4 instead of Multilingual v2.
  • Larger media uploads: Pro and higher plans can upload audio and video assets of up to 3 GB, up from 1 GB.
  • Word uploads: Audiobook projects can be created from Word .docx files as well as EPUB and PDF.
  • Video clip background removal: Video clips have a Remove background action, and right-clicking a clip opens its AI Tools.
  • Pronunciations tab: Audiobook projects have a pronunciations sidebar tab, opened from the selection menu or with Mod+Shift+P.
  • Fixes: The import speech endpoints for external audio and video clips return 429 system_busy and refund the transcription charge when transcription capacity is unavailable, instead of returning 500. Duplicating a project no longer fails when an image clip points at a source image that is not copied.
  • Close codes: The WebSocket always closes with a code: 1000 for a normal end, 1008 for a rejected request such as an unknown voice_id or an authentication failure, and 1011 for a server error.
  • Invalid frames: Frames that are not a JSON object receive an input_error message and the session continues.
  • Fixes: Translated audio is encoded in the requested output_format, so μ-law and A-law output plays correctly. Repeated utterances are no longer dropped, Chinese, Japanese and Korean clauses are no longer spoken twice or skipped, and Korean word order is improved.
  • SVG assets: SVG uploads to Assets stay SVG and render as vectors, including in the image editor. Scripts, event handlers and remote references are removed, models receive a PNG, and asset responses omit preview_content_url for SVGs.
  • Flows: Node groups and the structure sidebar are available in every workspace. LLM nodes on video-capable models such as gemini-3.5-flash-llm accept video prompt references.
  • Clearer reference errors: Generations whose reference file was refused on upload, or is not decodable audio, fail as a validation error with a clear message instead of an unexpected error.
  • Longer conversations: Speech Engine conversations can last up to 2 hours, up from 40 minutes. See Speech Engine.
  • Request headers: Speech Engine request header values can reference workspace secrets and are validated in the dashboard. Existing headers must be cleared before the WebSocket URL can change.
  • Larger voice selections: Manual voice selection is no longer limited to 10 voices, and link-shared voice collections offer Select all. Shared collection links open in the EU, India and Singapore data residency environments.
  • Fixes: The voice verification captcha accepts Serbian readings in Latin or mixed script.
  • Higher upload limit: Upload music accepts up to 100 songs per user per 24 hours, up from 20.
  • Architect permission: Architect access is controlled by its own workspace permission. PermissionType adds agents_architect and group permissions add architect, so admins can turn off Architect without removing Agents access, and API keys have a separate Architect scope.
  • History download permission: Download history items requires the speech_history_read API key permission.
  • Fixes: New users whose only invite is to a child workspace that inherits or borrows SSO can sign up with that SSO instead of receiving 400 provider_not_allowed.
  • Architect charging: Architect usage is now billed from workspace credits as its own line item, separate from agents’ minutes and LLM cost. Billing is rolling out gradually; the composer in the dashboard shows when it applies to your workspace. See Pricing and usage.
  • Realtime pricing: Workspace pricing tables list Realtime API audio (eleven_realtime_v1_audio) and the Eleven Realtime v1 Mini and Max token rates.
  • Cached LLM tokens: Cache writes are billed at least at the input rate and cache reads at the vendor’s cached rate. LLM calls for post-call analysis, guardrails and agent tests bill cached prompt tokens at those rates instead of the full input rate. Gemini 2.5 Flash-Lite cache reads are billed at 0.01 instead of 0.
  • Preview conversation discount: The discount on conversations started from the agent tester applies only to workspaces on 2022 pricing. On 2026 pricing, charging.dev_discount is false.
  • Fixes: OAuth discovery names the host the client connected to, so MCP clients such as Claude Code, Cursor and VS Code can connect to https://api.elevenlabs.io/v1/mcp.

JavaScript SDK

  • v2.76.0 - The realtime Speech to Text wrapper accepts entityRedaction. Added client.conversationalAi.analysisItems for managing analysis items, and methods on client.conversationalAi.realtime to list, create, get, update and delete realtime conversation groups and to set or clear a conversation’s group.

Python SDK

  • v2.76.0 - The realtime Speech to Text wrapper accepts entity_redaction. Added client.conversational_ai.analysis_items for managing analysis items, and list_conversation_groups, create_conversation_group, get_conversation_group, update_conversation_group, delete_conversation_group, set_conversation_group and clear_conversation_group on client.conversational_ai.realtime.

Android SDK

  • v0.13.1 - textOnly is now sent to the server, so text-only sessions run as text conversations instead of voice conversations. The top-level textOnly setting takes precedence over overrides.conversation.textOnly, as in the JavaScript SDK, and setting only the override now selects the WebSocket transport.

Unity SDK

  • v0.1.0 - First release of the Unity SDK for voice agents, installed through the Unity Package Manager from a git URL. It requires Unity 6.3 LTS or later and runs in the Editor, desktop players and WebGL, with WebRTC available on WebGL. Conversation.StartSessionAsync starts a session with an agent ID, signed URL or conversation token, and the SDK supports client tools, conversation events, messaging, feedback, device selection and spatial audio through an AudioSource. GettingStarted and QuickStart samples are included.

ElevenLabs CLI

  • v1.9.0 - agents workflow diagram draws an agent workflow in the terminal, or prints it as Mermaid with --format mermaid or --format mermaid-md. agents graph shows how agents, tools, tests, knowledge base documents and handoffs connect across the agents in agents.json, and --uses lists the agents that reference a given tool or document. Both commands read local files only. Regenerated against the latest API.
  • v1.8.0 - Request User-Agent headers include how the CLI was installed (install/<source>, such as npx, Homebrew or the installer). Pass --telemetry off, or set ELEVENLABS_TELEMETRY=off or DO_NOT_TRACK=1, to drop these tags. Regenerated against the latest API.

Packages

  • @elevenlabs/client@1.34.0, @elevenlabs/types@0.28.0 and @elevenlabs/react@1.21.0 - Scribe.connect() and useScribe() accept entityRedaction and entityRedactionMode, and dispatch the redacted_transcript message after each committed transcript. useScribe() adds an onRedactedTranscript callback and sets redactedText on the matching entry in committedTranscripts. stripAudioTags removes audio tags written in any language, not only tags in English letters.
  • @elevenlabs/convai-widget-core@0.24.0 - With persistent-session enabled, a voice call can be resumed after a page navigation. The next page shows the conversation so far with “Resume talking?”, and resuming starts a new conversation that continues from those turns, skipping the greeting if the agent allows first message overrides.
  • @elevenlabs/client@1.32.0 - Self-hosted worklets (workletPaths) are checked against the installed client version. A worklet copied from another release fails at session start with an error that names the worklet and path. Re-copy the files from @elevenlabs/client/worklets/* when upgrading.
  • @elevenlabs/client@1.32.1 - Fixes the start of the agent’s first message playing too fast and too high when the browser cannot use the agent’s sample rate. Voice sessions wait for the output resampler and fail with an error if it cannot be created.
  • @elevenlabs/client@1.33.0, @elevenlabs/types@0.27.0 and @elevenlabs/convai-widget-core@0.23.0 - Added a cloneStrength TTS override to the client and an override-clone-strength attribute to the widget. History rows of resumed persistent text conversations can carry rich_content, and the widget replays it.
  • @elevenlabs/react@1.20.3 - useScribe with autoConnect keeps a single open session under React StrictMode.
  • @elevenlabs/convai-widget-core@0.22.0 - Added the transcript-speech-sync attribute; set it to false to show the agent’s full message at once in voice calls. Carousel cards accept an optional rating from 0 to 5 and review_count.
  • @elevenlabs/convai-widget-core@0.21.6 and 0.22.2 - Voice calls show typing dots when the agent takes more than a second to start replying, or “Thinking…” under the avatar when the transcript is off. Partially revealed markdown links show their text instead of [blocked], carousel arrows no longer flicker, and carousel images use a 4:5 ratio.
  • @elevenlabs/react@1.20.1, 1.20.2, @elevenlabs/react-native@1.2.34, 1.2.35, 1.2.36, 1.2.37, @elevenlabs/convai-widget-core@0.22.1, 0.24.1 and @elevenlabs/convai-widget-embed@0.21.6 to 0.24.1 - Updated to the latest @elevenlabs/client and widget releases.

New endpoints

ElevenAgents

  • Analysis items
    • GET /v1/convai/analysis-items - List analysis items, with optional search, kind (evaluation or data_collection), analysis_item_ids, created_by_user_id, sort_direction, page_size and cursor.
    • POST /v1/convai/analysis-items - Create an analysis item.
    • GET /v1/convai/analysis-items/{analysis_item_id} - Get an analysis item.
    • PATCH /v1/convai/analysis-items/{analysis_item_id} - Update an analysis item.
    • DELETE /v1/convai/analysis-items/{analysis_item_id} - Delete an analysis item. Pass force=true to delete it even when agents use it, removing it from the dependent agents and branches.
    • POST /v1/convai/analysis-items/{analysis_item_id}/copy - Copy an analysis item.
    • GET /v1/convai/analysis-items/{analysis_item_id}/versions - List an analysis item’s versions.
    • GET /v1/convai/analysis-items/{analysis_item_id}/dependent-agents - List the agents that use an analysis item.
    • PATCH /v1/convai/analysis-items/{analysis_item_id}/attachments - Attach or detach an analysis item across agents and branches.
  • Realtime conversation groups
    • GET /v1/convai/realtime/conversation-groups and POST /v1/convai/realtime/conversation-groups - List and create conversation groups. Create requires title (string, unique per workspace, up to 64 characters) and accepts optional description (string, up to 500 characters).
    • GET, PATCH and DELETE /v1/convai/realtime/conversation-groups/{conversation_group_id} - Get, update and delete a conversation group.
    • PUT /v1/convai/realtime/conversations/{conversation_id}/conversation-group - Place a conversation in a group, with required conversation_group_id. A conversation belongs to at most one group.
    • DELETE /v1/convai/realtime/conversations/{conversation_id}/conversation-group - Remove a conversation from its group.

Updated endpoints

ElevenAgents

  • Create agent and the agent get and update endpoints
    • The prompt llm default changed from gemini-2.5-flash to gemini-3.5-flash.
    • TurnConfig adds noisy_environment_detection (boolean, default false) and noisy_environment_message (string). Language presets add nullable noisy_environment_message_translation, turn overrides and workflow overrides add nullable noisy_environment_message, and the override config adds noisy_environment_message (boolean, default false).
    • Widget configuration adds transcript_speech_sync_enabled (boolean, default true).
  • LLM enum adds gpt-6.1-terra, including in conversation_config_override on the /v1/convai/conversation WebSocket.
  • List conversations and Get conversation details - Added nullable user_message_count and agent_message_count (integer).
  • Get conversation details
    • WhatsApp conversation info adds nullable call_result (answered, voicemail, busy, no_answer or failed) and opening_message (template_name, wamid, sent_at_unix, delivered_at_unix, read_at_unix, replied_at_unix, failed_at_unix, error_code, pricing_category and billable).
    • WhatsApp call permission requests add nullable responded_at_unix (integer), permanent (boolean) and granted_by_user_call (boolean).
    • Agent definition source enum adds architect.
  • Update document - Added optional enable_auto_sync (boolean), auto_remove (boolean), minimum_frequency_days (integer) and tighten_only (boolean, default false).
  • Update triage ticket - Added optional qa_comment (string).
  • List test invocations - Added optional created_by_user_id query parameter (string).
  • Create test and the other test endpoints - Tool call evaluations add nullable guardrail_trigger (object with type any).
  • Rebase a branch onto main (POST /v1/convai/agents/{agent_id}/branches/{branch_id}/rebase) - The request body accepts application/json.

Speech to Text

  • Realtime (/v1/speech-to-text/realtime)
    • Added optional entity_redaction (string or array of strings) and entity_redaction_mode (redacted, entity_type or enumerated_entity_type, default enumerated_entity_type) query parameters.
    • Added the redacted_transcript server message.
    • session_started config adds entity_redaction (array of strings) and entity_redaction_mode (string).

Workspaces

  • Service account and workspace group permissions - PermissionType adds agents_architect, and workspace group permissions add architect.

Fast LLM service tier

ElevenAgents can now run supported LLMs on a fast service tier, which trades a per-token price premium for higher token throughput and lower latency. Set service_tier to fast in the agent’s prompt configuration with Update agent, and check which tiers a model supports in the available_service_tiers field returned by List LLMs. The default remains standard, and the LLM usage calculator accepts service_tier to estimate the cost difference.

Knowledge base crawl exclusions

ElevenAgents knowledge base website crawls can now skip parts of a site. Pass up to 50 exclude_patterns to Create crawl job, or fill in the Exclude patterns field in the whole-website crawl dialog, and URLs that match any pattern are skipped even when they match the crawl pattern. Crawl job responses return the patterns that were used. See Knowledge base for crawl setup.

Clone strength for v4 voices

Text to Speech and Text to Dialogue requests now accept clone_strength, a value from 0 to 1 that controls how closely a Professional Voice Clone fine-tuned for a v4 model follows its training samples. Set it in voice_settings on Create speech or in settings on Create dialogue. The recommended value for most use cases is 0. On the Text to Dialogue endpoints, clone_strength now decides whether a v4 PVC runs as a PVC or an IVC, and use_pvc_as_ivc is deprecated.

Expiring service account API keys

Workspaces service account API keys can now expire. Pass expires_at_unix when you create or update a key, or pick an Expire after value in the dashboard, and the key stops working after that time. Expiries must be between 15 minutes and 90 days in the future; omit the field for a key that never expires. See Service accounts.

All changes

  • Fast LLM service tier: Agents and workflow overrides accept service_tier (standard or fast) on supported models. List LLMs returns available_service_tiers for each model, and the LLM usage calculator accepts service_tier.
  • Knowledge base crawl exclusions: Create crawl job accepts exclude_patterns, and the whole-website crawl dialog adds an Exclude patterns field. Matching URLs are skipped.
  • Procedures analytics: The Procedures tab on the agent analytics dashboard is available to every workspace. Its Procedure invocations chart shows how often each procedure runs over time, grouped by procedure or by agent, and its tooltip shows whether a procedure is structured or free-form.
  • Start procedures at conversation start: Conversation initiation client data accepts start_procedure_ids, up to five free-form procedures that start when the conversation starts. The agent must have enable_procedure_ids_from_client enabled; structured procedures and unknown IDs fail the conversation start.
  • Structured procedure steps: Structured procedures can use the end_procedure system tool as a step, and the End step is relabeled End conversation. Sub-procedure and end steps now compile to system tool nodes, so agents start and end the intended procedure more reliably.
  • Channel integrations out of alpha: The Slack, Telegram, Custom Channel and Intercom integrations are no longer in alpha. Freshdesk remains in alpha.
  • Slack thread replies: The Slack Channel Message trigger adds a Respond to setting (response_mode), so an agent can keep replying in a thread after a single mention. Mentioning the agent with only stop, mute or quiet silences it in that thread until it is mentioned again, and events that Slack delivers twice get one reply.
  • Test run names: Run tests accepts an optional name, which test invocations and their summaries return. PATCH /v1/convai/test-invocations/{test_invocation_id} renames a run, and setting name to null restores the generated title.
  • Branch descriptions: Branch descriptions can be edited from the branches table with Update branch (description, up to 4,096 characters) and appear in the version tooltip and version graph.
  • Agent duplication keeps traffic splits: Duplicating an agent copies its live traffic split across the copied branches. The resulting deployment records duplicate as its deployment source.
  • Delete environment variables: DELETE /v1/convai/environment-variables/{env_var_id} deletes an environment variable.
  • Widget fullscreen on mobile: Widget configuration adds mobile_fullscreen (default true), with a Fullscreen on mobile switch in the dashboard. Set it to false to keep the compact layout on phones. See Widget.
  • Issues: Triage is renamed Issues in the agents dashboard, and triage tickets are now called issues. Every QA comment and flag_issue_for_review call opens its own issue, and an issue can be reopened while the conversation has another open issue. The API endpoints keep their triage-tickets paths. See Triage.
  • Prompt caching guidance: The system prompt editor flags system variables that change on every turn and prevent prompt caching, and Spotlight adds an Improve prompt caching recommendation for affected agents.
  • Realtime voice mode: Switching an agent to Realtime voice mode in the dashboard now selects Eleven Realtime v1 Max. Realtime agents always run tool calls in parallel, so the Parallel tool calling toggle is removed from Realtime model settings. See Realtime models.
  • Analytics: Workflow analytics can be filtered by channel, and the Guardrails tab is shown to all users.
  • Analysis LLMs: claude-sonnet-5-5 can be selected as the analysis LLM for evaluation criteria and data collection, both for the agent and per item. Agents on Claude Sonnet 5.5 can also set reasoning effort to none.
  • Conversation cost: Conversation charging adds voice_surcharge_charge, the credits charged as the voice actor reward for the Voice Library voices a conversation used. It is not included in cost.
  • Call duration and billing: Call charges and stored call duration stop when the call ends, so teardown time after the call is no longer billed or counted. Calls no longer continue for up to 50 seconds after ending when a knowledge base lookup is in progress.
  • Telephony: SIP trunk calls forward the User-To-User (UUI) header to the conversation initiation webhook. API keys with an IP allowlist now authenticate on the Genesys, Avaya, AudioCodes and Amazon Connect connector WebSockets. Batch call recipients whose call cannot be placed are marked failed instead of staying dispatched.
  • Observability: OpenTelemetry post-call webhook agent turn spans add elevenlabs.rag.chunks and elevenlabs.rag.used_chunk_ids, matching the JSON post-call webhook. See OpenTelemetry traces.
  • Guardrails: In voice conversations, the prompt injection guardrail now evaluates the user turn being answered instead of the agent’s previous message. Conversations on asynchronous channels such as Zendesk, Intercom and Custom Channel that a guardrail blocks now report the guardrail as the termination reason.
  • Zero Retention Mode: Synthetic voice detection is skipped for conversations in Zero Retention Mode.
  • WhatsApp: WhatsApp conversations resumed by a reply to a batch call template keep their batch_call metadata and batch ID.
  • Phone number secrets: Workspace secrets used as a Twilio auth token or an Exotel API token on a phone number show as in use and cannot be deleted while the number depends on them.
  • Fixes: Text conversations end as soon as the end_call tool runs instead of waiting for the next user message. Voice agents no longer go silent when a turn is cancelled before any audio is sent. The soft timeout no longer fires right after the first message or after tool calls on Eleven v4 voices. Realtime agents with sub-agent, agent tool or prompt say nodes that inherit the LLM can be saved and published. Agent tools that use OAuth2 connections no longer fail with expired tokens when the provider rotates refresh tokens. The nemotron-35-super-vl-nvfp4 context limit is raised from 64,000 to 131,072 tokens.
  • Clone strength: clone_strength (0 to 1) in Text to Speech voice_settings and Text to Dialogue settings controls how closely a Professional Voice Clone fine-tuned for a v4 model follows its samples. The recommended value for most use cases is 0. Edit voice settings stores clone_strength and Get voice settings returns it.
  • PVC and IVC on v4 dialogue: On the Text to Dialogue endpoints, clone_strength now decides whether a v4 PVC runs as a PVC or an IVC, and use_pvc_as_ivc is deprecated.
  • New endpoint paths: Voice Changer is documented at POST /v1/voice-changer/{voice_id} and POST /v1/voice-changer/{voice_id}/stream. The /v1/speech-to-speech/{voice_id} paths keep working but are deprecated and removed from the API reference. The Python and JavaScript SDKs call the new paths from v2.75.0, and the SDK method names are unchanged for now. In the next major SDK version, client.speechToSpeech will be deprecated in favor of client.voiceChanger to match the endpoint. See Voice changer.
  • Expiring service account API keys: Service account API keys accept expires_at_unix on create and update, between 15 minutes and 90 days in the future. Later expiries are rejected with 400 invalid_api_key_expiry. On update, pass "clear" to remove an expiry.
  • Group concurrency limits enforced: The workspace group concurrency limits (group_concurrency_limit, group_dubbing_concurrency_limit and group_music_concurrency_limit) are enforced for Text to Speech, Voice Changer, Music and Dubbing requests.
  • Voice cloning permissions: Members without the Create instant voice clone or Create professional voice clone permission can no longer create clones, even with general voice access.
  • SCIM group seat types: Workspaces that use SCIM without SCIM routing can assign a seat type to SCIM-synced groups from the Groups page.
  • Fixes: Delete invite returns 404 invite_not_found for an invite that does not exist, instead of a 500 error.
  • Shorter realtime silence thresholds: Realtime Speech to Text honors vad_silence_threshold_secs values below 0.5 seconds, which were previously treated as 0.5. The minimum accepted value is lowered from 0.3 to about 0.19 seconds.
  • Webhook retries: Transcription and error webhooks are retried on failure when the webhook’s retry setting is on. See Speech to Text webhooks.
  • Legacy endpoints deprecated: The /v1/dubbing endpoints outside /v1/dubbing/project, including Dub a video or an audio file and List dubs, are marked deprecated in the API reference.
  • Transcript downloads: The dubbing project menu can download target language transcripts as JSON, and as WebVTT subtitles for recent dubs.
  • Media folders: Workspace library files can be organized into folders in the Studio media browser.
  • Video clip actions: Video clips offer voice changer and voice isolator actions.
  • Video to music: Studio’s video to music workflow uses Music v2.5 by default.
  • Retired models: Image & Video and Flows requests for a retired model fail before generation with a 410 model_retired error that names the replacement model. Flows preflight reports the error on the affected node.
  • Fixes: Generations wait for a just-uploaded reference asset to finish processing instead of failing.
  • Fixes: Cantonese Professional Voice Clone samples are no longer rejected when language detection reports Chinese, and adding PVC samples no longer fails with a 500 error when language detection is unavailable.
  • Minimum charge: Text to Speech and Speech to Text requests priced below one credit are charged one credit, and refunds return exactly the credits that were charged.
  • Business plan image and video pricing: Business plans on the current pricing pay the same image and video credit rate as other self-serve plans, about 17% fewer credits per image and per video second than before.
  • OAuth scope: OAuth clients can request the realtime_translation scope.

JavaScript SDK

  • v2.75.0 - client.speechToSpeech.convert and client.speechToSpeech.stream now call POST /v1/voice-changer/{voice_id} and its /stream path. Added client.environmentVariables.delete, client.conversationalAi.tests.invocations.update for renaming a test invocation, and an optional name on client.conversationalAi.agents.runTests. Also adds serviceTier on the LLM usage calculators, excludePatterns on knowledge base crawl jobs, expiresAtUnix on service account API keys, cloneStrength in voice settings, startProcedureIds in conversation initiation client data and the widget mobileFullscreen setting. The legacy dubbing list, create, get and delete methods are marked deprecated, and the branch rebase request type BranchesRebaseRequest is renamed to BodyRebaseABranchOntoMainV1ConvaiAgentsAgentIdBranchesBranchIdRebasePost.

Python SDK

  • v2.75.0 - Added the callback_disconnect option to Conversation and AsyncConversation. It receives a DisconnectionDetails with a reason of "user", "agent" or "error", plus the WebSocket close code and reason when available, and replaces the deprecated callback_end_session. The regenerated client moves client.speech_to_speech to the /v1/voice-changer paths and adds client.environment_variables.delete, client.conversational_ai.tests.invocations.update and the same new parameters as JavaScript SDK v2.75.0.

Android SDK

  • v0.13.0 - agent_tool_request is now handled as a notification event instead of a client tool call, so client tools no longer run twice, the first time without parameters. The SDK also no longer crashes when a client tool answers after the connection has closed.

ElevenLabs CLI

  • v1.7.1 - say uses eleven_v4 and new agent templates use eleven_v4_turbo by default. Fixes onboard sign-in on Windows and when the token cannot be saved to the keychain.

Packages

  • @elevenlabs/client@1.29.0 and @elevenlabs/react@1.19.0 - Added the onAudioAlignmentPlayback callback, called with { event_id, alignment } when an aligned audio chunk starts playing, for syncing text with speech. It is not called on WebRTC connections.
  • @elevenlabs/client@1.30.0 - Added the opt-in stripAudioTags session option, which removes audio tags such as [laughs] from agent messages passed to onMessage and onAgentResponseCorrection, and exports the stripAudioTags helper. Stripped text no longer lines up with onAudioAlignment timings.
  • @elevenlabs/client@1.31.0, @elevenlabs/types@0.26.0 and @elevenlabs/react@1.20.0 - Added overrides.tts.modelId for a per-conversation TTS model override. Realtime Speech to Text emits RealtimeEvents.AUDIO_STARTED once the first microphone audio frame is sent, and useScribe in microphone mode reports connected only from that point.
  • @elevenlabs/convai-widget-core@0.21.0 and @elevenlabs/convai-widget-embed@0.21.0 - Voice call transcripts are revealed in step with the agent’s speech, and an interrupted reply keeps only what was said. Resuming a persistent text conversation no longer asks the user to accept the terms again.
  • @elevenlabs/convai-widget-core@0.21.1, 0.21.3 and 0.21.4 - Text chat shows typing dots while the agent prepares its reply. Rich content carousels show a page indicator, stay clickable after the conversation moves on, and appear in voice calls while the agent is still speaking. On screens narrower than 768px the chat opens fullscreen; add mobile-fullscreen="false" to the embed to keep the compact layout.
  • @elevenlabs/react-native@1.2.31, 1.2.32 and 1.2.33 - Updated to the latest @elevenlabs/client and @elevenlabs/react releases.

New endpoints

ElevenAgents

  • DELETE /v1/convai/environment-variables/{env_var_id} - Delete an environment variable. Returns 204.
  • PATCH /v1/convai/test-invocations/{test_invocation_id} - Rename a test invocation with required name (string, 1 to 256 characters, or null to restore the generated title).

Voice Changer

  • Voice changer - POST /v1/voice-changer/{voice_id} and Voice changer stream - POST /v1/voice-changer/{voice_id}/stream - The documented paths for Voice Changer. POST /v1/speech-to-speech/{voice_id} and POST /v1/speech-to-speech/{voice_id}/stream are removed from the API reference and deprecated.

Updated endpoints

ElevenAgents

  • Create agent, Get agent and Update agent
    • Prompt configuration adds service_tier (standard or fast, default standard). Workflow node prompt overrides add nullable service_tier.
    • Widget configuration adds mobile_fullscreen (boolean, default true).
  • List LLMs - Each model adds required available_service_tiers (array of standard or fast).
  • Calculate LLM usage and Calculate expected LLM usage for an agent - Added optional service_tier.
  • Get widget - Widget configuration adds mobile_fullscreen (boolean, default true).
  • Create crawl job - Added optional exclude_patterns (array of strings, up to 50 items of up to 2,048 characters each).
  • Get crawl job and List crawl jobs - Crawl jobs add exclude_patterns (array of strings).
  • Knowledge base document responses - auto_sync_info removes the deprecated auto_discover field.
  • Run tests - Added optional name (string, 1 to 256 characters). Get test invocation and List test invocations return nullable name.
  • Update branch - Added optional description (string, up to 4,096 characters). Pass an empty string to clear it.
  • Get conversation details
    • Conversation charging adds nullable voice_surcharge_charge (integer).
    • Agent definition source enum adds architect.
  • Agent deployments (GET /v1/convai/agents/{agent_id}/deployments) - Deployment source enum adds duplicate.
  • Conversation initiation client data, including the conversation_initiation_client_data WebSocket event and Submit batch call - Added optional start_procedure_ids (array of strings).
  • Tool type enum removes api_integration_mcp.

Text to Speech and Text to Dialogue

Dubbing

  • List dubs, Dub a video or an audio file, GET /v1/dubbing/{dubbing_id}, DELETE /v1/dubbing/{dubbing_id}, GET /v1/dubbing/{dubbing_id}/audio/{language_code} and GET /v1/dubbing/{dubbing_id}/transcripts/{language_code}/format/{format_type} - Marked deprecated.

Workspaces


Eleven Realtime

Eleven Realtime is a new family of models for ElevenAgents voice agents, rolling out gradually across workspaces. Each one is a complete voice agent configuration. You pick one model, and ElevenLabs handles the LLM, Speech to Text and Text to Speech settings behind it. Your agent’s system prompt, voice, tools and knowledge base work the same as before.

There are two to choose from. Eleven Realtime v1 Mini (eleven_realtime_v1_mini) is the default, and it’s the fastest and cheapest. Eleven Realtime v1 Max (eleven_realtime_v1_max) reasons before it answers, so use it when your agent has complex instructions or multi-step tool calls.

If you’ve already built on OpenAI’s Realtime API, you don’t need to rewrite your client. The events and the WebSocket transport are the same, so you change the endpoint and the session configuration. Your backend calls Create realtime client secret to get a short-lived secret, and the browser uses it to connect directly.

Realtime Translation

Realtime Translation turns speech in one language into speech in another while the speaker is still talking. It’s a WebSocket API, available now for enterprise workspaces.

You open a connection with a target_language and a voice_id, then stream audio in. You get translated speech back in that voice. You don’t have to say which language the speaker is using, because it’s detected automatically, but you can set source_language if you know it. Along the way, the server also sends live transcripts of the original speech and the translated text, for example to show captions.

eagerness controls the tradeoff between speed and quality. high starts translating soonest, and low waits for more of the sentence. The translated voice uses Eleven v4 Turbo by default, and you can switch to eleven_v3 or eleven_flash_v2_5 with tts_model_id. For sensitive conversations, set enable_zrm=true to use Zero Retention Mode.

Video to Music

Video to Music writes a soundtrack for your video. The Video to music API is now available: you upload a video and get back music that fits it, without writing a music prompt from scratch. Use it to score product videos, ads and social clips.

You can send up to 10 video files (200 MB combined). They’re joined in order and scored as one video. To steer the result, add a text description and up to 10 style tags.

Three options are worth knowing. use_audio_as_reference takes the genre, BPM, instruments and mood from the video’s own audio. add_sfx adds sound effects that match the video. separate_tracks returns the music and the sound effects as separate files in a ZIP, so you can mix them yourself.

ElevenAgents

  • Alerts: Alerts notify a workspace when agent error rates exceed a threshold or rise sharply above their recent baseline. Monitors cover conversation failures, tool failures, webhook errors, guardrail triggers, call initiation errors and evaluation success scores, and run with default thresholds without any setup. Agent-level thresholds and notifiers can be set through platform_settings.alerting on Update agent. Notifications can be delivered as webhooks, to Slack or to PagerDuty.
  • Guardrails generally available: Guardrails are no longer in alpha.
  • Workspace environments: Environments for environment variables can now be managed through the API with List environments, Create environment and Delete environment. A variable without a value for an environment falls back to its production value. Deleting an environment removes its values from all variables, and production cannot be deleted.
  • Answering machine detection actions: Twilio answering machine detection settings on Outbound call and Submit batch call accept actions, a list of rules that pair an action (hang_up) with the answered_by verdicts that trigger it, such as machine_start or fax. Verdicts without a rule leave the call running.
  • New LLMs: Agents can use gemini-4-argon, claude-opus-4-6 and the ElevenLabs-hosted nemotron-35-super-vl-nvfp4. See Models.
  • Authored-only workflow reads: Get agent and Update agent accept include_procedure_generated_workflow_elements. Set it to false to receive only the workflow nodes and edges you authored, excluding the ones compiled from structured procedures. Defaults to true, so existing callers are unchanged.
  • Branch rebasing from any branch: Rebase branch and Preview rebase accept an optional source_branch_id to pull changes from a branch other than main.
  • Delete agents used as subagents: Delete agent accepts force. When true, the agent is deleted even if other agents use it as a subagent, and it is removed from those agents and their branches.
  • Sort agents by last update: List agents accepts sort_by=updated_at, and each agent summary includes updated_at_unix_secs, the last time the main branch or agent-level settings changed.
  • Google Calendar event updates: The Google Calendar integration adds google_calendar_update_event and google_calendar_delete_event tools. Both are privileged tools that can change or cancel any event on the connected calendar.

Music and Sound Effects

  • Zero Retention Mode: Music generation, composition plan, stem separation and video to music endpoints, plus Sound Effects, accept enable_logging=false to use Zero Retention Mode. Zero Retention Mode is available to enterprise customers. Music requests that set store_for_inpainting=true are rejected in this mode.
  • WAV output for sound effects: Sound Effects accepts wav_8000, wav_16000, wav_22050, wav_24000, wav_32000, wav_44100 and wav_48000 output formats.

Speech Engine

  • 40-minute conversations: Speech Engine supports conversations of up to 40 minutes. The default limit remains 10 minutes. Raise conversation.max_duration_seconds to allow longer conversations.
  • Sort by last update: List Speech Engines accepts sort_by=updated_at, and each summary includes updated_at_unix_secs.

Dubbing

  • Subtitles for dubbed languages: Dubbing project language outputs (GET /v1/dubbing/project/{project_id}/language/{language_id}) include subtitles, a signed URL to the translated lines as WebVTT subtitles timed to the source speech. The URL expires one hour after it is issued.

Voices

  • Resubmittable voice reviews: Voice sharing review_status adds declined_resubmittable for declined voices that can be resubmitted for review.

Workspaces

  • Disable service account API keys: Disable service account API key disables a service account API key without changing its other settings. It remains available after a workspace downgrades from a plan that includes service accounts.
  • Group concurrency limits: List workspace groups and List service accounts return group_concurrency_limit, group_dubbing_concurrency_limit and group_music_concurrency_limit for each group, and group permissions add all_resources_access.

SDK releases

JavaScript SDK

  • v2.74.0 - Added client.conversationalAi.realtime.createClientSecret for the Realtime API, with the RealtimeModel and OpenAI-compatible session configuration types, and an environments client with list, create and delete. Also adds force on agent deletion, sourceBranchId on branch rebase and rebase preview, includeProcedureGeneratedWorkflowElements on agent reads, enableLogging on Music and Sound Effects methods, WAV output formats for Sound Effects, service account API key disable, and Twilio answering machine detection actions.
  • v2.73.0 - Added createAndWait helpers for Flows image, video and Text to Speech generations and template runs (client.flows.templates.runs.createAndWait). They create the request and poll until it completes or fails, pacing requests with the server’s Retry-After header. A failed result is returned rather than thrown, and waitTimeoutInSeconds throws FlowsWaitTimeoutError if the work is not finished in time. The regenerated client adds the Video to music add_sfx, separate_tracks and use_audio_as_reference options.

Python SDK

  • v2.74.1 - music.compose_detailed now accepts and forwards enable_logging, which the custom wrapper previously dropped, so Zero Retention Mode works with this method.
  • v2.74.0 - Added client.conversational_ai.realtime.create_client_secret for the Realtime API and an environments client, with the same regenerated API changes as JavaScript SDK v2.74.0.
  • v2.73.0 - Added create_and_wait helpers for Flows image, video and Text to Speech generations and template runs, for both sync and async clients, with wait_timeout_in_seconds and min_poll_interval_in_seconds options and a FlowsWaitTimeoutError exported from elevenlabs.flows_custom. Type checkers such as pyright now resolve client.webhooks.construct_event, client.speech_to_text.realtime, client.text_to_speech.convert_realtime and the custom Music methods. Includes the same regenerated API changes as the JavaScript SDK.
  • v2.73.1 - Fixed the test suite on Python 3.13 and later, and added Python 3.13 and 3.14 to CI.

ElevenLabs CLI

  • v1.7.0 - Regenerated API commands, adding environments, realtime client secrets and service account API key disabling.
  • v1.6.0 - Added create-and-wait commands for flows image, flows video, flows text-to-speech and flows templates runs. Each command returns a finished result, with --wait-timeout and --min-poll-interval options.
  • v1.5.1, v1.5.2, v1.5.3 and v1.6.1 - Bug fixes for onboarding, including Windows support.

Packages

API

New endpoints

ElevenAgents

  • Create realtime client secret - POST /v1/convai/realtime/client_secrets - Create an ephemeral client secret for browser-based Realtime API sessions. Accepts an optional session (OpenAI-compatible session configuration with model of eleven_realtime_v1_mini (default) or eleven_realtime_v1_max, instructions, audio, tools, turn_detection and max_duration_seconds from 60 to 7,200, default 600) and optional expires_after with required seconds (integer). Returns value (string), expires_at (integer) and the echoed session.

  • List environments - GET /v1/convai/environments - List the workspace environments, including production. Each environment has name, nullable updated_at_unix_secs and nullable created_by_user_id.

  • Create environment - POST /v1/convai/environments - Create an environment with a required, unique name (string). Returns 409 if the name is taken.

  • Delete environment - DELETE /v1/convai/environments/{environment} - Delete an environment and remove its values from all environment variables. Returns 204.

Workspaces

  • Disable service account API key - POST /v1/service-accounts/{service_account_user_id}/api-keys/{api_key_id}/disable - Disable a service account API key without changing its other settings.

Updated endpoints

ElevenAgents

  • Get agent, Update agent, Preview rebase and Preview merge
    • Added optional include_procedure_generated_workflow_elements (boolean, default true). When false, the returned workflow contains only authored nodes and edges, without those generated from structured procedures.
  • Create agent, Get agent and Update agent
    • The llm enum adds gemini-4-argon, claude-opus-4-6 and nemotron-35-super-vl-nvfp4.
    • TTS configuration effects removes the deprecated seed field.
  • Delete agent - Added optional force (boolean, default false).
  • List agents - sort_by adds updated_at. Agent summaries add required updated_at_unix_secs (integer).
  • Rebase branch and Preview rebase - Added optional source_branch_id (string). Defaults to the agent’s main branch.
  • Outbound call, Submit batch call and Get batch call
    • Twilio machine detection configuration adds actions, an array of rules with required action (hang_up) and required answered_by (array of human, fax, unknown, machine_start, machine_end_beep, machine_end_silence or machine_end_other). Each action can appear in at most one rule.
  • Get conversation details
    • Transcript tool calls add nullable tool_id (string). It is not set for tools without an ID, such as system tools.
    • Turn metrics add nullable voice_id (string).
    • Tool results add nullable to_agent and to_node (string).
    • WhatsApp conversation info adds nullable recipient with nullable phone and bsuid (string), and nullable call_permission_request with required state (blocked, sent, accepted or declined) and nullable requested_at_unix (integer).
    • The platform charge category enum adds v3_addon, v4_addon and rag_addon.
  • Get ticket, List tickets and List workspace tickets
    • Ticket responses add required status_changes, an array of objects with status, nullable changed_by_user_id (string) and changed_at_unix_secs (integer).

Text to Speech and Text to Dialogue

Music

Sound Effects

  • Create sound effect
    • Added optional enable_logging query parameter (boolean, default true).
    • output_format adds wav_8000, wav_16000, wav_22050, wav_24000, wav_32000, wav_44100 and wav_48000.

Speech Engine

  • List Speech Engines - sort_by adds updated_at. Summaries add required updated_at_unix_secs (integer).

Speech to Text

  • Create transcript and Get transcript - Edited transcripts add required, read-only edited_text (string), matching text and the edited_text field of the realtime edited_transcript event.

Dubbing

  • GET /v1/dubbing/project/{project_id}/language and GET /v1/dubbing/project/{project_id}/language/{language_id} - Language outputs add nullable subtitles (string, signed URL).

Voices

Workspaces

  • List workspace groups and List service accounts
    • Groups add nullable group_concurrency_limit, group_dubbing_concurrency_limit and group_music_concurrency_limit (integer or unlimited).
    • Group permissions add all_resources_access.

ElevenAgents

  • Merge and search triage tickets: Merge tickets folds up to 50 open tickets from the same agent into one target ticket. Merged tickets keep the status merged and reference the target through merged_into_ticket_id. List tickets and List workspace tickets add a case-insensitive search parameter that matches ticket titles, descriptions, comments and turn comments, and each result includes a search_match excerpt showing where the query matched.
  • Workflow End node outcomes: Workflow End nodes accept an outcome of success (default) or failure, so a workflow can record whether a conversation ended successfully. The conversation error type adds workflow_end_failure.
  • Call queueing enabled for new agents: New agents are created with call queueing enabled unless platform_settings.queueing_config.enabled is set explicitly. Existing agents keep their current setting.
  • Post-call audio webhook retries: When Enable retries is on, post_call_audio webhooks are retried on the same schedule as transcription webhooks. Audio payloads over 50 MiB are not retried, and queued audio retries are capped at 400 MiB per webhook.
  • Strict tool overrides: The conversation override tool_ids_strict makes an overridden tool_ids list the complete set of tools available in the conversation, including tools on workflow nodes. It is ignored when set on the agent itself.
  • Widget end confirmation: The widget can ask users to confirm before ending a chat or call. Enable it with end_confirmation_enabled in the widget configuration or the end-confirmation attribute.

Text to Speech

  • Audio effects: Text to Speech and Text to Dialogue requests accept an optional audio_effects object. It applies a filter preset, perceived distance, a reverb environment, a looped background noise bed and stereo panning to the generated audio. List valid preset, environment and background noise IDs with GET /v1/audio-effects/catalog. Reverb and non-zero panning produce stereo audio and are rejected for ulaw_*, alaw_* and pcm_* output formats. Effects do not change the cost of a request.

Speech to Text

  • Turn prediction commit strategy: The realtime Speech to Text WebSocket adds turn_prediction as a commit_strategy value. In addition to silence-based segmentation, it commits a segment as soon as the model predicts the speaker has finished their turn. It is the default for models that predict turns, and partial transcripts from those models include turn_probability. See transcripts and commit strategies.

Voices

  • Voice collections API: Voice collections can now be managed through the API. Create, list, update and delete collections, and add or remove voices. See List voice collections.

Workspaces

  • Basic seat fallback: Update member accepts allow_downgrade_to_basic_seat. When unlocking a user whose original seat type has no seats available, the user is moved to a basic seat instead of the request failing.
  • Workspace ID on user: Get user returns the workspace_id of the workspace the API key belongs to.

SDK releases

JavaScript SDK

  • v2.72.0 - Added clients for voice collections, the audio effects catalog and triage ticket merging. Text to Speech and Text to Dialogue methods accept audio_effects, triage ticket list methods add search and filters, and the realtime Speech to Text client adds the turn_prediction commit strategy.

Python SDK

  • v2.72.0 - Added clients for voice collections, the audio effects catalog and triage ticket merging. Text to Speech and Text to Dialogue methods accept audio_effects, triage ticket list methods add search and filters, and the realtime Speech to Text client adds the turn_prediction commit strategy. play() and save() now accept any iterable of audio chunks, such as a list or tuple. RealtimeEvents.CLOSE handlers receive a dictionary with the WebSocket close code and reason; handlers without parameters continue to work.

ElevenLabs CLI

  • v1.5.0 - Regenerated API commands, adding voice collections, the audio effects catalog, triage ticket merging and agent merge proposals.

Packages

  • @elevenlabs/client@1.27.0 and @elevenlabs/react@1.17.0 - Added a previousText option to Scribe.connect() and useScribe() in microphone mode. The text is sent with the first audio chunk to give the model context, such as existing document text, for casing, punctuation and sentence continuation. MediaDeviceOutput now receives libsampleratePath, so output resampling uses the same custom libsamplerate bundle path as MediaDeviceInput.
  • @elevenlabs/react-native@1.2.29 - Updated to @elevenlabs/client@1.27.0 and @elevenlabs/react@1.17.0.
  • @elevenlabs/convai-widget-core@0.19.0 and @elevenlabs/convai-widget-embed@0.19.0 - Added an optional confirmation dialog before ending a chat or call, enabled with the end-confirmation attribute or end_confirmation_enabled config, with customizable text. Added carousel rich content with scrollable item cards that can include an image, subtitle, description and actions. Carousels remain in the transcript after a reply, without their actions. Rich content is now shown during voice calls when the transcript is visible.

API

New endpoints

ElevenAgents

  • Merge tickets - POST /v1/convai/triage-tickets/{agentqa_ticket_id}/merge - Merge open tickets into the target ticket. Requires source_ticket_ids (string array, 1–50 items) and returns the target ticket.

Voices

  • List voice collections - GET /v1/voices/collections - List the voice collections the caller can access, sorted by title.
  • Create voice collection - POST /v1/voices/collections - Create a collection with required title and icon (string, maximum 1,000 characters each).
  • Get voice collection - GET /v1/voices/collections/{collection_id} - Retrieve a collection with its title, icon and the caller’s permission_on_resource (admin, editor, commenter or viewer).
  • Update voice collection - PATCH /v1/voices/collections/{collection_id} - Update the optional title or icon.
  • Delete voice collection - DELETE /v1/voices/collections/{collection_id}
  • Add voice to collection - POST /v1/voices/collections/{collection_id}/voice/{voice_id}
  • Remove voice from collection - DELETE /v1/voices/collections/{collection_id}/voice/{voice_id}

Audio effects

  • GET /v1/audio-effects/catalog - List the available filter_presets, environments and background_noises. Each entry has an ID, a name and a nullable description.

Updated endpoints

ElevenAgents

  • Create agent, Get agent and Update agent
    • Workflow End nodes add outcome (success or failure, default success). The field is required in responses.
    • Standalone workflow agent nodes and agent transfer configurations add nullable branch_id (string) to target a specific branch of the destination agent. When unset, the transfer follows the target agent’s traffic deployment. Transfer tool results add a configured branch reason.
    • Prompt override configurations add tool_ids_strict (boolean, default false).
    • TTS configuration effects add pan (number, -1 to 1, default 0). The seed field is deprecated and has no effect.
  • Get widget config - Added end_confirmation_enabled (boolean, default false).
  • List tickets
    • Added optional search (string, maximum 200 characters) and merged_into_ticket_id (string) query parameters.
    • Ticket responses add nullable merged_into_ticket_id and nullable search_match with field (title, description, comment or turn_comment), snippet, highlight_start, highlight_end and nullable turn_index.
  • List workspace tickets - Added optional agent_id, sources, priorities, sort_by (created_at or priority, default created_at), owner_user_id, issue_type, label, merged_into_ticket_id and search query parameters.
  • List tests - Added optional created_by_user_id (string). Use @me for the authenticated user.
  • List test invocations - Added optional test_id (string), version_id (string), start_time and end_time (integer, Unix seconds, inclusive) query parameters.
  • Update agent branch - Added optional auto_rebase (boolean). Branch responses add auto_rebase (boolean, default false).
  • Get conversation details
    • Twilio, SIP trunk and Exotel phone call metadata add nullable call_result with answered, voicemail, busy, no_answer and failed values.
    • Batch call metadata adds nullable cold_attempt (integer) and nullable conversation_type (initial, retry, callback or redirect).
    • The conversation error type adds workflow_end_failure.
  • Knowledge base auto-sync auto_discover is deprecated and no longer has any effect.

Text to Speech

  • Create speech, Stream speech, Create speech with timing and Stream speech with timing
    • Added optional nullable audio_effects with filter_preset_id (string), distance (number, 0–1, default 0), environment_id (string), background_noise_id (string), send_level (number, 0–1, default 1) and pan (number, -1 to 1, default 0).
    • Unknown effect IDs return 404. Stereo effects return 400 for ulaw_*, alaw_* and pcm_* output formats.

Text to Dialogue

Speech to Text

  • Realtime Speech to Text
    • The commit_strategy query parameter adds turn_prediction, alongside manual and vad.
    • partial_transcript messages add optional turn_probability (number), sent only for models that predict turns.

Music

  • POST /v1/music/video-to-music - Responses add a song-id header with the ID of the generated song.

History

Workspace and user

  • Update member - Added optional allow_downgrade_to_basic_seat (boolean, default false).
  • Get user - Added required workspace_id (string).
  • Service account API key permissions add voice_collections_read and voice_collections_write.

ElevenAgents

  • 구조화된 절차는 저장 및 게시 시 컴파일됩니다: 이제 에이전트 업데이트 및 에이전트 초안 생성이 구조화된 절차를 자동으로 컴파일하고 검증합니다. 더 이상 통합에서 컴파일 엔드포인트를 호출하거나 생성된 워크플로를 수동으로 업데이트할 필요가 없습니다. 유효하지 않은 절차는 400 procedure_validation_failed로 절차별 오류를 반환합니다. 구조화된 절차가 없는 에이전트는 영향을 받지 않습니다.
  • 절차 컴파일 엔드포인트는 레거시입니다: 절차 컴파일은 레거시로 표시되며 새 통합에서 사용하면 안 됩니다. 기존 호출자에게는 계속 제공됩니다.
  • 병합 제안: 병합 제안은 알파 버전으로 제공됩니다. 팀은 에이전트 브랜치에 대한 제안을 열고, 구성 변경 사항과 테스트 결과를 검토하며, 댓글이나 승인 결정을 남기고, 승인된 변경 사항을 대상 브랜치에 병합할 수 있습니다. 병합 제안을 참조하세요.

SDK 릴리스

JavaScript SDK

  • v2.71.0 - 에이전트 병합 제안, 배포 기록, 테스트 호출 취소, Speech Engine 복제용 클라이언트를 추가했습니다. 또한 트리아지 티켓 우선순위 및 정렬, 배치 통화 내보내기 제한, gpt-6.1-sol, claude-sonnet-5-5, Eleven v4 에이전트 음성 모델 및 최신 에이전트 구성 타입을 추가했습니다.

Python SDK

  • v2.71.0 - 에이전트 병합 제안, 배포 기록, 테스트 호출 취소, Speech Engine 복제용 클라이언트를 추가했습니다. 또한 트리아지 티켓 우선순위 및 정렬, 배치 통화 내보내기 제한, gpt-6.1-sol, claude-sonnet-5-5, Eleven v4 에이전트 음성 모델 및 최신 에이전트 구성 타입을 추가했습니다.

Swift SDK

  • v3.4.0 - 이제 동적 변수가 대화 시작 데이터로 전송될 때 네이티브 JSON 타입을 유지합니다.

패키지

  • @elevenlabs/client@1.26.0, @elevenlabs/react@1.16.0, @elevenlabs/react-native@1.2.28, @elevenlabs/types@0.24.0 - transcriptEdit, edited_transcript 이벤트, React의 onEditedTranscript 콜백을 통해 실시간 음성-텍스트 변환 트랜스크립트 편집을 추가했습니다. 이제 React는 filterBackgroundAudio를 실시간 세션으로 전달합니다. 클라이언트는 MessagePayload도 내보내며, 중복 에이전트 답변을 방지하기 위해 안정적인 응답 ID를 사용합니다.
  • @elevenlabs/convai-widget-core@0.18.3 및 @elevenlabs/convai-widget-embed@0.18.3 - 첨부 파일만 있는 답변을 포함해 에이전트 메시지 첨부 파일의 이미지 미리 보기와 다운로드 링크를 추가했습니다. 안정적인 응답 ID가 중복 답변을 방지하며, 이제 텍스트 전용 대화에서는 채팅 종료 툴팁이 있는 중지 컨트롤을 사용합니다.

API

새 엔드포인트

ElevenAgents

  • 병합 제안 생성 - POST /v1/convai/agents/{agent_id}/merge-proposals - 필수 source_branch_id, target_branch_id, title 필드와 선택적 설명, 검토자 ID, 트리아지 티켓 ID로 제안을 생성합니다.
  • 병합 제안 나열 - GET /v1/convai/agents/{agent_id}/merge-proposals - 상태, 소스 브랜치, 대상 브랜치, 텍스트 필터 및 커서 페이지네이션으로 제안을 나열합니다.
  • 병합 제안 가져오기 - GET /v1/convai/agents/{agent_id}/merge-proposals/{merge_proposal_id} - 제안과 검토, 댓글, 결과 세부 정보를 조회합니다.
  • 병합 제안 업데이트 - PATCH /v1/convai/agents/{agent_id}/merge-proposals/{merge_proposal_id} - 제목, 설명 또는 요청된 검토자를 업데이트하거나 병합하지 않고 닫습니다.
  • 병합 제안 검토 - POST /v1/convai/agents/{agent_id}/merge-proposals/{merge_proposal_id}/reviews - 선택적 댓글과 함께 approved 또는 changes_requested 결정을 제출합니다.
  • 병합 제안에 댓글 달기 - POST /v1/convai/agents/{agent_id}/merge-proposals/{merge_proposal_id}/comments - 최대 4,096자의 Markdown 댓글을 추가합니다.
  • 병합 제안 병합 - POST /v1/convai/agents/{agent_id}/merge-proposals/{merge_proposal_id}/merge - 소스 브랜치를 병합하고, 필요에 따라 강제 충돌 해결 및 소스 브랜치 보관을 제어합니다.
  • 에이전트 배포 나열 - GET /v1/convai/agents/{agent_id}/deployments - 페이지 기반 페이지네이션으로 현재 및 과거 브랜치 트래픽 분할을 나열합니다.
  • 테스트 호출 취소 - POST /v1/convai/test-invocations/{test_invocation_id}/cancel - 실행 중인 테스트 호출을 취소합니다.

Speech Engine

  • Speech Engine 복제 - POST /v1/speech-engine/{speech_engine_id}/duplicate - Speech Engine을 복제하고, 선택적으로 사본에 새 name을 할당합니다.

업데이트된 엔드포인트

ElevenAgents

  • 에이전트 생성, 에이전트 가져오기, 에이전트 업데이트
    • LLM 열거형에 gpt-6.1-sol 및 claude-sonnet-5-5를 추가했습니다.
    • 대화형 TTS 모델 옵션에 eleven_v4 및 eleven_v4_turbo를 추가했습니다. 기본값은 eleven_flash_v2에서 eleven_v4_turbo로 변경되었습니다.
    • 에이전트 전송 및 독립형 워크플로 에이전트 노드에 전송 후 기본 음성, TTS 모델 및 음성 설정을 유지하는 선택적 preserve_voice_settings(불리언, 기본값 false)를 추가했습니다.
    • 턴 건너뛰기 도구 구성에 선택적 wait_timeout_secs(숫자, 기본값 -1)를 추가했습니다. 값이 -1이면 발신자가 말할 때까지 에이전트가 무음 상태를 유지합니다.
    • 통화 후 웹훅 구성에 JSON 트랜스크립션 웹훅에서 턴별 트랜스크립트를 생략하는 선택적 exclude_transcript(불리언, 기본값 false)를 추가했습니다.
    • 대화 기록 수정 구성에 근거를 수정하는 동안 선택한 추출 값을 유지하는 선택적 excluded_data_collection_ids(문자열 배열, 최대 50개)를 추가했습니다.
    • 이제 클라이언트 도구 response_timeout_secs는 expects_response가 활성화된 경우 무기한 대기하도록 -1을 허용합니다.
  • 티켓 나열
    • low, medium, high, urgent 값을 포함하는 선택적 priorities(배열)를 추가했습니다.
    • created_at 및 priority 값을 갖는 sort_by를 추가했으며, 기본값은 created_at입니다.
    • 이제 티켓 응답에는 title, nullable priority, priority_changes 배열이 필요합니다. 티켓 생성 및 업데이트 요청에서 priority를 설정할 수 있으며, 티켓 상태에 cancelled가 추가되었습니다.
  • 도구 나열 - 에이전트의 기본 브랜치에서 참조되는 도구를 반환하는 선택적 nullable used_by_agent_id(문자열)를 추가했습니다.
  • 배치 통화 결과 내보내기 - 첫 번째 수신자만 열 미리 보기로 내보내는 선택적 nullable limit(정수, 최소 1)를 추가했습니다.
  • 테스트 호출 나열 및 테스트 호출 가져오기
    • 테스트 호출 요약에 필수 cancelled_count(정수) 및 선택적 cancelled(불리언, 기본값 false)를 추가했습니다.
    • 테스트 실행 상태에 cancelled를 추가했습니다.
  • 대화 세부 정보 가져오기 - 트랜스크립트 항목에 user_turn_timeout 및 skip_turn_wait_elapsed 값을 갖는 선택적 nullable platform_event를 추가했습니다.
  • 에이전트 버전 메타데이터 가져오기 - 기본 브랜치에 병합된 변경 사항을 게시한 사람을 설명하는 배열인 선택적 nullable merged_authors를 추가했습니다.
  • 크롤 작업 가져오기 - 지원 중단된 max_depth 응답 필드를 제거했습니다.

Eleven v4 및 Eleven v4 Turbo

이제 Eleven v4 및 Eleven v4 Turbo를 사용할 수 있습니다. Eleven v4는 90개 이상의 언어에서 더 높은 품질과 표현력이 풍부한 음성, 향상된 음성 복제를 제공합니다. Eleven v4 Turbo는 실시간 애플리케이션을 위한 동일한 모델 제품군으로, 중앙값 추론 지연 시간은 약 100ms입니다.

콘텐츠 제작 및 장문 오디오에는 Text to Dialogue API를 통해 eleven_v4를 사용하세요. 에이전트 및 인터랙티브 애플리케이션에는 Text to Dialogue WebSocket을 통해 eleven_v4_turbo를 사용하세요.

음성-텍스트 변환

  • 트랜스크립트 편집: 이제 배치 및 실시간 트랜스크립션에서 최대 2,000자의 자연어 편집 지시문을 받을 수 있습니다. 배치 응답은 edited_transcript에 결과를 반환하며, 실시간 세션은 edited_transcript 이벤트를 내보냅니다. 호환되지 않는 기능, 응답 처리 및 가격은 배치 및 실시간 가이드를 참조하세요.

ElevenAgents

  • 에이전트 모델: 지원되는 에이전트 LLM 옵션에 gpt-6-sol, gpt-6-luna, glm-52, deepseek-v41-flash, claude-opus-5, claude-opus-5-5를 추가했습니다.
  • 에이전트 테스트 사용량: 테스트 호출 요약에 총 크레딧 및 USD 가격을 포함할 수 있습니다. 개별 테스트 실행에는 크레딧 및 과금 내역을 포함할 수 있습니다.

Flows

  • 템플릿 API: 게시된 Flows 템플릿을 나열하고 검색하며, 템플릿 실행을 시작하고 실행 결과를 조회하는 엔드포인트를 추가했습니다. 실행은 게시된 버전을 대상으로 할 수 있으며, 최종 결과를 구성된 웹훅으로 전송할 수 있습니다.

Text to Dialogue

  • 생성 연속성: Text to Dialogue 요청에서 이전 또는 이후 텍스트와 요청 ID를 사용해 여러 생성 간 연속성을 유지할 수 있습니다.
  • 프로페셔널 음성 모드: Text to Dialogue 요청에서 프로페셔널 음성 복제의 즉석 음성 복제 버전을 사용해 지연 시간을 줄이거나 표현력을 높일 수 있습니다.

SDK 릴리스

JavaScript SDK

  • v2.70.0 - 배치 및 실시간 음성-텍스트 변환 트랜스크립트 편집 타입, 테스트 호출 요약용 branchId 필터링 및 비용 필드, 지식 베이스 자동 검색, Text to Dialogue usePvcAsIvc, deepseek-v41-flash 에이전트 모델을 추가했습니다.

Python SDK

  • v2.70.0 - 배치 및 실시간 음성-텍스트 변환 트랜스크립트 편집 타입, 테스트 호출 요약용 branch_id 필터링 및 비용 필드, 지식 베이스 자동 검색, Text to Dialogue use_pvc_as_ivc, deepseek-v41-flash 에이전트 모델을 추가했습니다.

ElevenLabs CLI

  • v1.4.0 - 명시적 API 키를 사용하는 요청이 저장된 OAuth 자격 증명도 함께 전송하여 인증에 실패하던 문제를 수정했습니다. 피드백 명령 그룹에 설명을 추가하고, 생성된 에이전트 스킬에서 누락된 기능을 신고하는 방법을 추가했습니다. Flows 템플릿 엔드포인트 및 최신 스키마 타입으로 API 클라이언트를 다시 생성했습니다.

API

새 엔드포인트

Flows

  • 템플릿 나열 - GET /v1/flows/templates - 게시된 템플릿과 실행 가능한 버전을 나열합니다.
  • 템플릿 가져오기 - GET /v1/flows/templates/{template_id} - 템플릿과 실행 가능한 버전을 조회합니다.
  • 템플릿 실행 생성 - POST /v1/flows/templates/{template_id}/runs - 입력 값, 선택적 버전 및 선택적 웹훅으로 템플릿 실행을 시작합니다.
  • 템플릿 실행 나열 - GET /v1/flows/templates/{template_id}/runs - 커서 페이지네이션 및 선택적 버전 필터링으로 템플릿 실행을 나열합니다.
  • 템플릿 실행 가져오기 - GET /v1/flows/templates/{template_id}/runs/{run_id} - 템플릿 실행과 출력 상태를 조회합니다.

업데이트된 엔드포인트

음성-텍스트 변환

  • 음성을 텍스트로 변환 - POST /v1/speech-to-text
    • 최대 2,000자의 자연어 편집 지시문을 받을 수 있는 선택적 nullable transcript_edit(문자열)를 추가했습니다.
    • 응답에 선택적 nullable edited_transcript를 추가했습니다. 편집에 성공하면 필수 kind: "transcript" 및 text 필드가 포함됩니다. 편집에 실패하면 필수 kind: "error", error_type: "edit_failed", message 필드가 포함됩니다.
  • 실시간 음성-텍스트 변환은 transcript_edit(문자열)를 수락하고 필수 text 및 edited_text 필드가 있는 edited_transcript 이벤트를 내보냅니다.

ElevenAgents

  • 에이전트 생성, 에이전트 가져오기, 에이전트 업데이트
    • LLM 열거형에 gpt-6-sol, gpt-6-luna, glm-52, deepseek-v41-flash, claude-opus-5, claude-opus-5-5를 추가했습니다.
    • 알림 모니터에 모니터의 알림 전송 여부를 제어하는 선택적 nullable enabled(불리언)를 추가했습니다.
  • 대화 나열 및 텍스트 검색
    • data_collection_params 및 dynamic_variable_params에 neq 및 in 연산자를 추가했습니다. in의 값은 파이프 구분 기호를 사용합니다.
    • evaluation_params는 결과 값으로 success, failure 또는 unknown을 허용합니다.
  • 스마트 검색 - GET /v1/convai/conversations/messages/smart-search
    • 선택적 nullable branch_id(문자열)를 추가했습니다.
  • 테스트 호출 나열 - GET /v1/convai/test-invocations
    • 선택적 nullable branch_id(문자열)를 추가했습니다.
    • 테스트 호출 요약에 선택적 nullable credits_used(정수) 및 total_price(숫자)를 추가했습니다.
  • 테스트 호출 가져오기 - GET /v1/convai/test-invocations/{test_invocation_id}
    • 개별 테스트 실행에 선택적 nullable credits_used(정수) 및 charging(ConversationChargingCommonModel)을 추가했습니다.
  • 대화 시뮬레이션 및 대화 스트림 시뮬레이션
    • 선택적 new_turns_limit 기본값을 10,000에서 100으로 변경하고 최대값을 100으로 설정했습니다.
  • 크롤 작업 생성 및 크롤 작업 가져오기
    • 자동 동기화 중 링크를 따라가 새로 발견된 페이지를 추가하는 선택적 auto_discover(불리언, 기본값 false)를 추가했습니다. 이를 사용하려면 enable_auto_sync: true가 필요합니다.
  • 지식 베이스 문서 새로 고침 - POST /v1/convai/knowledge-base/{documentation_id}/refresh
    • AutoSyncInfo에 새로 고침 시 새로 발견된 페이지를 크롤링하고 추가하는지 나타내는 선택적 auto_discover(불리언, 기본값 false)를 추가했습니다.
  • MCP 서버 도구 나열 - GET /v1/convai/mcp-servers/{mcp_server_id}/tools
    • 응답에 선택적 tool_approval_statuses(배열)를 추가했습니다. 각 항목에는 필수 tool_id(문자열)와 state(up_to_date, needs_review 또는 not_approved)가 포함되며, 선택적 승인 정책 및 승인된 정의 필드가 포함됩니다.

Text to Dialogue

텍스트 음성 변환

Flows

보이스 디자인

워크스페이스

  • 서비스 계정 API 키 생성 - POST /v1/service-accounts/{service_account_user_id}/api-keys
    • 엔터프라이즈 서비스 계정 API 키를 위한 선택적 nullable tts_concurrency_limit, music_concurrency_limit, dubbing_concurrency_limit(정수)를 추가했습니다.
  • 서비스 계정 API 키 업데이트 - PATCH /v1/service-accounts/{service_account_user_id}/api-keys/{api_key_id}
    • 선택적 tts_concurrency_limit, music_concurrency_limit, dubbing_concurrency_limit를 추가했습니다. 각 필드는 정수, clear, no_update, null을 허용하며 기본값은 no_update입니다.

모델

  • 모델 나열 - GET /v1/models
    • 필수 max_characters_request_free_user 및 max_characters_request_subscribed_user를 지원 중단했습니다. 이 필드는 적용되지 않으므로 대신 maximum_text_length_per_request를 사용하세요.

이미지 & 비디오

  • Sora 2 및 Sora 2 Pro 지원 종료: OpenAI는 2026년 9월 24일 Sora API를 중단합니다. 두 모델은 이미지 & 비디오 모델 선택기에서 제거되었으며, 해당 날짜에 생성을 중단합니다. 기존 생성물은 기록에서 계속 사용할 수 있습니다. Sora 노드를 사용하는 Flows 및 템플릿은 다시 실행하려면 Gemini Omni 1.1 Flash와 같은 다른 비디오 모델로 전환해야 합니다.
  • Seedance 1.5 Pro 지원 중단: ByteDance는 2026년 11월 11일 Seedance 1.5 Pro를 종료합니다. 이 모델은 더 이상 새 생성에 제공되지 않으며 해당 날짜에 작동을 중단합니다. 대신 Seedance 2.0 모델을 사용하세요.

ElevenAgents

  • 병렬 도구 호출: 이제 에이전트 프롬프트 구성에 enable_parallel_tool_calls(불리언, 기본값 true)가 포함됩니다. 활성화하면 지원되는 모델은 한 턴에서 여러 도구를 실행할 수 있습니다.
  • 에이전트 모델: 에이전트 LLM 옵션에 gpt-6-astra를 추가했습니다.
  • 알림: 이제 에이전트 알림에서 PagerDuty 및 webhook 알림과 함께 Slack 채널을 지원합니다.
  • 전화번호 검색: 제공업체, 발신 지원, 할당된 에이전트 및 브랜치, 라벨, 전화번호 필터를 갖춘 커서 페이지네이션 전화번호 엔드포인트를 추가했습니다.

이미지 및 비디오

  • GPT Image 2.5 모델: 이제 이미지 생성에서 gpt-image-2.5-flare 및 gpt-image-2.5-sunburst를 지원합니다. 두 모델 모두 최대 10개의 참조 이미지, max까지의 품질 수준, auto를 포함한 14개의 고정 가로세로비, 1K, 2K 또는 4K 출력을 지원합니다.

Speech Engine

  • 캐스케이드 시간 초과: 이제 Speech Engine 생성 및 업데이트 요청에서 cascade_timeout_seconds(숫자, 2–15, 기본값 4)를 지원하여 재시도 전 ElevenLabs가 업스트림 음성 엔진을 대기하는 시간을 제어할 수 있습니다.

텍스트 음성 변환

  • 멀티 컨텍스트 턴 경계: 이제 is_final_audio_for_turn 마커가 MP3 및 Opus 출력을 포함해 해당 턴의 버퍼링된 모든 바이트 뒤에 전송됩니다. 클라이언트는 PCM으로 전환하지 않고도 이 마커를 정확한 턴 경계로 사용할 수 있습니다.

SDK 릴리스

JavaScript SDK

  • v2.69.0 - 실시간 Speech to Text 래퍼에 transcriptEdit를 추가했으며, 편집된 텍스트는 edited_transcript 이벤트를 통해 전달됩니다. 데이터 레지던시 접미사가 있는 API 키에 대한 Speech Engine JWT 검증을 수정했습니다. 템플릿을 나열하고 가져오며 실행을 나열, 생성, 가져오는 phoneNumbers.listV2 및 Flows 템플릿 메서드를 추가했습니다. GPT Image 2.5, Slack 에이전트 알림, 병렬 도구 호출, gpt-6-astra, 서비스 계정 API 키 동시성 제한, Music 제로 보존, 대화 필터, MCP 도구 승인 상태를 위한 클라이언트와 타입을 재생성했습니다.

Python SDK

  • v2.69.0 - 실시간 Speech to Text 래퍼에 transcript_edit를 추가했으며, 편집된 텍스트는 edited_transcript 이벤트를 통해 전달됩니다. 데이터 레지던시 접미사가 있는 API 키에 대한 Speech Engine JWT 검증을 수정하고, ffplay가 실패하면 play가 오류를 발생시키도록 했습니다. 템플릿을 나열하고 가져오며 실행을 나열, 생성, 가져오는 phone_numbers.list_v2 및 Flows 템플릿 메서드를 추가했습니다. GPT Image 2.5, Slack 에이전트 알림, 병렬 도구 호출, gpt-6-astra, 서비스 계정 API 키 동시성 제한, Music 제로 보존, 대화 필터, MCP 도구 승인 상태를 위한 클라이언트와 타입을 재생성했습니다.

ElevenLabs CLI

  • v1.3.2 - CLI 사용에 대한 피드백을 수집할 수 있도록 선택적 --intent 메타데이터를 추가했습니다. CLI에서 지원하지 않는 워크플로를 신고할 수 있도록 elevenlabs feedback missing-capability를 추가했습니다.

API

새 엔드포인트

ElevenAgents

  • 전화번호 페이지 나열 - GET /v1/convai/v2/phone-numbers
    • 선택적 page_size(정수, 1–1,000, 기본값 100) 및 cursor(nullable 문자열) 페이지네이션 매개변수를 추가했습니다.
    • 선택적 search, label, phone_number, provider, supports_outbound, agent_id, branch_id, sort_by, sort_direction 필터를 추가했습니다. sort_by는 label 또는 phone_number를 지원하며, sort_direction의 기본값은 asc입니다.
    • 응답에는 필수 phone_numbers(배열), 선택적 nullable next_cursor(문자열), has_more(불리언, 기본값 false)가 포함됩니다.

업데이트된 엔드포인트

ElevenAgents

  • 에이전트 생성, 에이전트 가져오기 및 에이전트 업데이트
    • 프롬프트 구성에 enable_parallel_tool_calls(불리언, 기본값 true)를 추가했습니다. 워크플로 재정의는 nullable 불리언을 지원합니다.
    • LLM 열거형에 대화 시작 재정의를 포함하여 gpt-6-astra를 추가했습니다.
    • 이제 알림 통합 알리미에서 integration_type을 사용하여 pagerduty 또는 slack을 선택합니다. Slack 알리미에는 connection_id 및 channel_id(문자열)가 필요하며, channel_id는 1~64자를 지원합니다.
  • 대화 나열 및 대화 참조 확인
    • data_collection_params에서 더 이상 missing 연산자를 지원하지 않습니다. 필드가 수집되지 않은 대화를 일치시키려면 빈 값을 사용하세요.
    • dynamic_variable_params의 빈 값은 변수가 설정되지 않은 대화와 일치합니다.
  • Speech Engine 생성, Speech Engine 가져오기 및 Speech Engine 업데이트
    • cascade_timeout_seconds(숫자, 2–15)를 추가했습니다. 생성 요청의 기본값은 4이며, 업데이트 요청은 nullable 값을 지원하고 응답에는 해당 필드가 필요합니다.

이미지 및 비디오

  • 이미지 생성 만들기 - POST /v1/flows/image
    • gpt-image-2.5-flare 및 gpt-image-2.5-sunburst를 위한 요청 변형을 추가했습니다.
    • 두 변형 모두 prompt 및 model_id가 필요하고, 최대 10개의 images를 지원하며, resolution(1K, 2K 또는 4K, 기본값 1K)을 추가합니다.
    • quality는 low, medium, high, xhigh, max를 지원하며 기본값은 high입니다. aspect_ratio는 auto 및 14개의 고정 비율을 지원하며 기본값은 16:9입니다.

더빙

지식 베이스

  • 크롤 작업 생성 - POST /v1/convai/knowledge-base/crawl
    • max_depth(정수, 기본값 3)를 지원 중단했습니다. 이 필드는 이제 아무 작업도 수행하지 않으며 최소 또는 최대 제약 조건도 აღარ 이상 적용되지 않습니다.
    • 크롤 작업 메타데이터에서 더 이상 max_depth가 필요하지 않습니다.

사용자

스키마 변경 사항

텍스트 음성 변환

  • 이제 멀티 컨텍스트 WebSocket audio 및 final 메시지에서 is_final_audio_for_turn을 해당 턴의 버퍼링된 모든 바이트가 전달된 후 발생하는 것으로 정의합니다.

ElevenAgents

  • 통화 대기열 및 대기 오디오: 동시 처리 한도에 도달한 에이전트를 위한 통화 대기열을 추가했습니다. 대기 중인 발신자는 대기 오디오를 듣고, 입장하거나 시간 초과될 때까지 queue_status 이벤트를 수신합니다. 새 API 메서드로 사용자 지정 대기 오디오를 업로드하거나 삭제할 수 있습니다.
  • 테스트 버전 메타데이터: 이제 에이전트 테스트 결과에서 테스트를 실행한 에이전트 버전과 초안 변경 사항 포함 여부를 확인할 수 있습니다. 호출 요약에는 부분 재제출이 서로 다른 버전에서 실행된 경우도 표시됩니다.
  • 모델 및 계정 메타데이터: 이제 에이전트 LLM 구성에서 gemini-3.8-flash를 지원합니다. WhatsApp 계정 응답에서 해당 계정이 Cloud API를 사용하는지, 공존 가입 플로를 사용하는지 확인할 수 있습니다.

음악

  • Music v2.5 API 지원: 이제 음악 엔드포인트와 SDK에서 music_v2_5를 지원합니다. 작곡 계획 생성 청크는 최대 6,132자를 지원하며, 각 줄 최대 200자의 줄을 최대 30개까지 포함할 수 있습니다.

SDK 릴리스

JavaScript SDK

  • v2.68.0 - 에이전트 대기 오디오를 업로드하고 삭제하는 메서드를 추가했습니다. 통화 대기열, 프로시저 참조, 초안 브랜치 생성, RAG 쿼리 제한, 테스트 버전 메타데이터, gemini-3.8-flash, music_v2_5를 위한 타입을 재생성했습니다.

Python SDK

  • v2.68.0 - 에이전트 대기 오디오를 업로드하고 삭제하는 동기 및 비동기 메서드를 추가했습니다. 통화 대기열, 프로시저, 초안 브랜치 생성, RAG 쿼리 제한, 테스트 버전 메타데이터, gemini-3.8-flash, music_v2_5를 위한 타입을 재생성했습니다. 이제 compose_detailed 래퍼에서 생성된 모든 음악 모델 ID와 프롬프트 기반 및 청크 기반 작곡 계획을 모두 지원합니다.

iOS SDK

  • v3.3.1 - 에이전트 메시지를 event_id 대신 response_id로 조정하여 도구 호출 주변의 여러 응답을 보존하고, 스트리밍된 부분이나 수정 사항이 잘못된 메시지에 연결되지 않도록 했습니다.

ElevenLabs CLI

  • v1.3.0 - 생성된 API 명령에 선택적 --intent 메타데이터를 추가했습니다. CLI에서 지원하지 않는 워크플로를 신고할 수 있도록 elevenlabs feedback missing-capability를 추가했습니다.

패키지

  • @elevenlabs/convai-widget-core@0.18.2 및 @elevenlabs/convai-widget-embed@0.18.2 - 도구 및 프로시저 호출 후 중복되는 에이전트 답변을 수정했습니다. 스트리밍된 텍스트 답변은 Markdown 형식을 유지하며, 음성 대화에서는 발화되지 않은 스트리밍 채팅 부분이 더 이상 렌더링되지 않습니다. 임베드 패키지는 user_activity 이벤트도 초당 한 번으로 제한하고 세션이 종료되면 이 제한을 해제합니다.

API

새 엔드포인트

ElevenAgents

  • 에이전트 대기 오디오 업로드 - POST /v1/convai/agents/{agent_id}/hold-audio
    • 필수 hold_audio_file(바이너리)을 multipart 양식 데이터로 받습니다. 파일은 최대 40MB 및 180초 길이의 MP3 또는 WAV여야 합니다.
    • 필수 agent_id(문자열) 및 hold_audio(AgentHoldAudioConfig)를 반환합니다. 대기 오디오 객체에는 audio_path, audio_url, original_filename(문자열), duration_secs(숫자), size_bytes(정수)가 필요합니다.
  • 에이전트 대기 오디오 삭제 - DELETE /v1/convai/agents/{agent_id}/hold-audio
    • 사용자 지정 클립을 제거하여 대기 중인 발신자가 기본 대기음을 듣도록 합니다.
    • 필수 agent_id(문자열)를 반환합니다.

업데이트된 엔드포인트

ElevenAgents

  • 에이전트 생성, 에이전트 가져오기 및 에이전트 업데이트
    • enabled(불리언, 기본값 false) 및 wait_timeout_seconds(정수, 1–1,800, 기본값 180)가 포함된 선택적 platform_settings.queueing_config(AgentQueueingConfig)를 추가했습니다.
    • 대기열 구성 응답에는 읽기 전용의 nullable hold_audio(AgentHoldAudioConfig)가 포함됩니다. 에이전트 응답에는 선택적 default_hold_audio_url(문자열)도 포함됩니다.
  • 대화 가져오기 - GET /v1/convai/conversations/{conversation_id}
    • 메타데이터에 선택적 nullable queue_wait_secs(숫자)를 추가했습니다. 대기열 시간은 call_duration_secs 및 청구 시간에서 제외됩니다.
  • 에이전트 브랜치 생성 - POST /v1/convai/agents/{agent_id}/branches
    • 선택적 include_draft(불리언, 기본값 false)를 추가했습니다. true인 경우 브랜치 팁에서 생성된 브랜치에 호출자의 초안 프로시저 세트가 포함됩니다.
  • 프로시저 나열 및 프로시저 생성, 가져오기, 초안 엔드포인트
    • 목록 항목에 선택적 referenced_tool_ids, referenced_kb_ids, referenced_procedure_ids, referenced_dynamic_variables(문자열 배열)를 추가했습니다.
    • 에이전트 및 병합 미리보기 응답에 프로시저 ID를 키로 하는 맵인 선택적 procedures를 추가했습니다.
  • 테스트 호출 나열 - GET /v1/convai/test-invocations
    • 이름으로 테스트와 폴더를 필터링하는 선택적 nullable search(문자열)를 추가했습니다.
  • 에이전트에서 테스트 실행 및 테스트 호출 가져오기
    • 테스트 스위트 호출 및 단위 테스트 실행 응답에 선택적 nullable version_id(문자열)와 선택적 ran_against_draft(불리언, 기본값 false)를 추가했습니다.
    • 호출 요약에도 선택적 runs_diverged_from_version(불리언, 기본값 false)를 추가했습니다.
  • POST /v1/convai/agents/{agent_id}/knowledge-base/rag-query
    • 요청 본문에 쿼리별 재정의 값으로 선택적 nullable max_documents_length(정수, 1–50,000) 및 max_retrieved_rag_chunks_count(정수, 1–20)를 추가했습니다.
  • WhatsApp 계정 가져오기 - GET /v1/convai/whatsapp-accounts/{phone_number_id}
    • 선택적 account_type(WhatsAppAccountType, 기본값 cloud_api)을 추가했습니다. 값은 cloud_api 및 coexistence입니다.
  • 예상 LLM 사용량 계산 - POST /v1/convai/agent/{agent_id}/llm-usage/calculate
    • LLMUsageCalculatorLLMResponseModel에 필수 price_per_message(숫자)를 추가했습니다.

음악

  • MusicModelID에 POST /v1/music/video-to-music, POST /v1/music/plan, 음악 작곡, POST /v1/music/detailed, POST /v1/music/detailed/stream, POST /v1/music/stream, POST /v1/music/upload, POST /v1/music/finetunes용 music_v2_5를 추가했습니다.
  • GenerationChunk.text의 maxLength를 POST /v1/music/plan, POST /v1/music, POST /v1/music/detailed, POST /v1/music/detailed/stream, POST /v1/music/stream, POST /v1/music/upload에 대해 6,000자에서 6,132자로 늘렸습니다. 섹션 이름은 1~100자를 포함할 수 있고, 청크에는 한 줄당 200자씩 최대 30줄을 포함할 수 있습니다.

스키마 변경 사항

ElevenAgents

  • 실시간 서버 이벤트에 queue_status를 추가했습니다. 필수 queue_status_event.status는 waiting, admitted, timed_out를 지원하며, timed_out은 WebSocket 종료 코드 4300보다 먼저 발생합니다. 대기 오디오는 계속 일반 audio 이벤트로 전달됩니다.
  • LLM 열거형에 대화 시작 재정의를 포함하여 gemini-3.8-flash를 추가했습니다.
  • 대화 시스템 도구 결과 유니온에 start_procedure_success, start_procedure_error, end_procedure_success, end_procedure_error를 추가했습니다.
  • FeaturesUsageCommonModel에 선택적 freeform_procedure 및 structured_procedure 기능 상태를 추가했습니다.