> This is a page from the ElevenLabs documentation. For a complete page index, fetch https://elevenlabs.io/docs/llms.txt. For the full documentation in a single file, fetch https://elevenlabs.io/docs/llms-full.txt. # Image & Video ## Overview Image & Video enables you to create high-quality visual content from simple text descriptions or reference images. Generate static images or dynamic videos in any style, then refine them iteratively with additional prompts, upscale for high-resolution output, and even add lip-sync with audio. > **Warning** > > Some Image & Video models and input upload capabilities are restricted in the United States due to > regulatory or provider requirements. Check individual model descriptions for specific > availability. > **Info** > > Free plan users can only generate images and are limited to three image requests per day. Video > generation requires a paid plan. Generating through the API requires a Pro plan or above. For API > requests, ByteDance models are disabled by default and require explicit approval before use. > Enterprise customers can contact support to request access. #### [Products](/docs/eleven-creative/playground/image-video) Complete guide to using Image & Video in ElevenLabs. #### [Developers](/docs/eleven-api/guides/cookbooks/image-and-video) Step-by-step guide for using Image & Video with the API. ## Key capabilities * **Image generation**: Create high-quality images from text prompts or reference images with models optimized for speed or quality * **Video generation**: Generate dynamic videos with cinematic motion, physics realism, and integrated audio. Video generation is only available on paid plans * **Iterative refinement**: Refine generations with additional prompts and create variations * **Enhancement tools**: Upscale resolution by up to 4x and apply realistic lip-sync with audio * **Multiple models**: Access specialized models for different use cases, from rapid iteration to production-ready content * **Reference support**: Guide generation with start frames, end frames, and style references. Supports a wide range of image file formats including JPG, PNG, WEBP, and more * **Export flexibility**: Download as standalone files or import directly into ElevenCreative Studio projects ## Workflow The creation process moves you from inspiration to finished asset in four stages: **Explore:** Discover community creations to find inspiration and study effective prompts. **Generate:** Use the prompt box to describe what you want to create, select a model, and fine-tune settings. **Iterate and enhance:** Review generations, create variations, and apply enhancements like upscaling and lip-syncing. **Export:** Download finished assets or send them directly to ElevenCreative Studio. ## Supported download formats **Video:** * **MP4**: Codecs H.264, H.265. Quality up to 4K (with upscaling) **Image:** * **PNG**: High-resolution, lossless output ## Models Image & Video provides access to specialized models optimized for different use cases. Each model offers unique capabilities, from rapid iteration to production-ready quality. Post-processing models require an existing generated output, though you can also upload your own image or video file. > **Info** > > Enterprise workspace admins can control which image and video generation models are available to > workspace members. By default, all models are disabled for Enterprise workspaces and must be > explicitly enabled by admins. Learn more about [Model approvals](/docs/overview/administration/workspaces/model-approvals). > **Warning** > > For API requests, ByteDance models are disabled by default and require explicit approval before > use. Enterprise customers can contact support to request access. > **Info** > > Every model below is available in the Image & Video app. Models that can also be called from the > [Image & Video API](/docs/eleven-api/guides/cookbooks/image-and-video) list their `model_id` under > **API**; the rest are app-only. #### Video generative models #### Seedance 2.0 A unified multimodal video model with audio-video joint generation, offering director-level control over performance, lighting, shadow, and camera movement for ultra-realistic, cinematic results. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References * Video References * Audio References **Features:** * Unified multimodal architecture that jointly generates synchronized audio and video in a single pass * Industry-leading multimodal reference and editing capabilities, accepting text, image, audio, and video inputs simultaneously * Exceptional motion stability with cinematic-grade realism aligned to industry standards * Director-level creative control over performance, lighting, shadow, and camera movement * Flexible generation lengths from 4s up to 15s * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p, 1080p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Cinematic storytelling requiring synchronized audio and visuals * Reference-driven content with strict adherence to source images, videos, or audio * Professional productions requiring fine-grained control over lighting and camera work **Cost:** Varies based on selected settings and duration **API:** [`bytedance-seedance-v2`](/docs/api-reference/flows/video/create#request.body.bytedance-seedance-v2) > **Note** > > Settings can be toggled to adjust credit consumption. > **Warning** > > Seedance 2.0 is not available in the United States. #### Seedance 2.0 Fast A faster, lower-cost variant of Seedance 2.0 with the same multimodal reference inputs, capped at 720p output. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 9) * Video References (up to 3, 15s combined) * Audio References (up to 3, 15s combined) **Features:** * Same reference handling as Seedance 2.0, with a combined cap of 12 references per generation * Frames and references are mutually exclusive: use a start or end frame, or references, not both * Flexible generation lengths from 4s up to 15s * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Iterating on Seedance 2.0 prompts before a full-resolution render * Reference-driven clips where 720p output is sufficient **Cost:** Varies based on selected settings and duration **API:** [`bytedance-seedance-v2-fast`](/docs/api-reference/flows/video/create#request.body.bytedance-seedance-v2-fast) > **Warning** > > Seedance 2.0 Fast is not available in the United States. #### Seedance 2.0 Mini The smallest Seedance 2.0 variant: roughly twice as fast as Seedance 2.0 at about half the cost, with the same inputs and 720p output. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 9) * Video References (up to 3, 15s combined) * Audio References (up to 3, 15s combined) **Features:** * Same reference handling as Seedance 2.0, with a combined cap of 12 references per generation * Frames and references are mutually exclusive: use a start or end frame, or references, not both * Flexible generation lengths from 4s up to 15s * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * High-volume drafts and previews * Cost-sensitive workflows that still need audio and reference inputs **Cost:** Varies based on selected settings and duration **API:** [`bytedance-seedance-v2-mini`](/docs/api-reference/flows/video/create#request.body.bytedance-seedance-v2-mini) > **Warning** > > Seedance 2.0 Mini is not available in the United States. #### Seedance 2.5 The successor to Seedance 2.0 with sharper realism, up to 50 combined image, video, and audio references, and generations up to 30 seconds. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 30) * Video References (up to 10, 30s combined) * Audio References (up to 10, 30s combined) **Features:** * No combined cap across reference types; audio-only references are accepted without an image or video * Frames and references are mutually exclusive * Flexible generation lengths from 4s up to 30s * Aspect ratio is taken from the start frame or reference video when one is provided * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Long-form clips that need to stay consistent with many references * Dialogue and performance scenes driven by reference audio **Cost:** Varies based on selected settings and duration **API:** [`bytedance-seedance-v2.5`](/docs/api-reference/flows/video/create#request.body.bytedance-seedance-v2.5) > **Note** > > Seedance 2.5 does not expose a seed control. > **Warning** > > Seedance 2.5 is not available in the United States. #### Seedance 2.5 Video Edit Edits an existing video by describing the changes. Output length and aspect ratio follow the input video. **Generation inputs:** * Video to edit (required, up to 30s) * Text prompt describing the edits * Image References (up to 30) * Additional Video References (up to 10, 30s combined) * Audio References (up to 10, 30s combined) **Features:** * No duration control: the output matches the input video length * Aspect ratio is taken from the input video * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p **Ideal for:** * Changing wardrobe, props, lighting, or setting in existing footage * Style transfer that preserves the original motion **Cost:** Varies based on selected settings and duration > **Warning** > > Seedance 2.5 Video Edit is not available in the United States. #### Seedance 2.5 Video Extend Continues an existing video from where it ends, guided by a prompt and optional references. **Generation inputs:** * Video to extend (required, up to 30s) * Text prompt describing the continuation * Image References (up to 30) * Additional Video References (up to 10, 30s combined) * Audio References (up to 10, 30s combined) **Features:** * Extension length from 4s up to 30s * Aspect ratio is taken from the input video * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p **Ideal for:** * Lengthening a shot that ends too early * Chaining several generations into a longer sequence **Cost:** Varies based on selected settings and duration > **Warning** > > Seedance 2.5 Video Extend is not available in the United States. #### Kling 3.0 An advanced video model that functions like an AI director, maintaining high consistency for characters, items, and scenes across complex camera movements. **Generation inputs:** * Text-to-Video * Start Frame * End Frame **Features:** * High-fidelity character and scene retention using multi-angle image or video references * Native audio-visual co-generation with multilingual lip-sync and environmental sound * Flexible generation lengths from 3s up to 15s * Generate up to 4 variations simultaneously * Enhanced handling of text, fluid dynamics, and complex physical interactions **Output options:** * Resolutions: 1080p only * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * Character-driven storytelling requiring visual continuity * Commercials and assets with specific text-rendering needs **Cost:** Varies based on selected settings and duration > **Note** > > Supports negative prompts for granular control. Sound can be enabled or disabled per generation. > **Warning** > > Kling 3.0 is not available in the United States. #### Kling O3 A high-consistency video model that functions like an AI director, preserving the identity of characters, items, and scenes across complex camera movements. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Video Reference * Image Reference **Features:** * Maintains precise visual identity for main characters and items using multi-angle references * Supports seamless generation lengths from 3s up to 15s * Generate up to 4 variations at a time * Accurate modeling of element interactions and motion coherence * Native support for enabled or disabled audio per generation **Output options:** * Resolutions: 1080p only * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * Character-driven storytelling requiring strict visual continuity * Professional marketing and brand assets with consistent item rendering **Cost:** Varies based on selected settings and duration > **Note** > > Settings can be toggled to adjust credit consumption. > **Warning** > > Kling O3 is not available in the United States. #### Kling O3 Edit A natural language-driven video-to-video editing model based on the O3 architecture, enabling complex visual transformations—such as character replacement and environment swaps—without the need for manual masking or frame-by-frame adjustments. **Generation inputs:** * Source Video * Image References (up to 4 distinct elements/angles) * Text Description (Natural language instructions) **Features:** * Interprets conversational prompts to replace subjects or settings while respecting the original motion structure * Maintains original camera angles, movement patterns, and spatial relationships throughout the edit * Combines up to 4 total elements (including frontal and multi-angle images) to ensure high-fidelity character consistency * Option to preserve original source audio or generate silent output per generation * Generate up to 4 edited variations at a time **Output options:** * Resolutions: Dependent on source video * Aspect ratios: Matches source video **Ideal for:** * High-fidelity character replacement in existing footage while keeping original movements * Complete scene environment transformations (e.g., changing a daytime city to a futuristic nightscape) * Applying style transfers that require strict adherence to existing camera dynamics **Cost:** Varies based on video duration and selected settings > **Warning** > > Kling O3 Edit is not available in the United States. #### Google Veo 3.1 A professional-grade model for high-quality, cinematic video generation. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References **Features:** * Excellent quality and creative control with negative prompts * Fully integrated and synchronized audio * Realistic dialogue, lip-sync, and sound effects * Fixed durations: 4s, 6s, and 8s * Batch creation with up to 4 generations at a time * Dedicated sound control **Output options:** * Resolutions: 720p, 1080p * Aspect ratios: 16:9, 9:16 **Ideal for:** * High-quality, cinematic video generation with full creative control **Cost:** Varies based on selected settings and duration **API:** [`veo-3.1-generate-001`](/docs/api-reference/flows/video/create#request.body.veo-3.1-generate-001) > **Note** > > Enabling and disabling sound will change the generation credits. #### Google Veo 3.1 Fast A high-speed model optimized for rapid previews and generations, delivering sharper visuals with lower latency. **Generation inputs:** * Text-to-Video * Start Frame * End Frame **Features:** * Advanced creative control with negative prompts and dedicated sound control * Fixed durations: 4s, 6s, and 8s * Batch creation with up to 4 generations at a time * Accurately models real-world physics for realistic motion and interactions **Output options:** * Resolutions: 720p, 1080p * Aspect ratios: 16:9, 9:16 **Ideal for:** * Quick iteration and A/B testing visuals * Fast-paced social media content creation **Cost:** Varies based on selected settings and duration **API:** [`veo-3.1-fast-generate-001`](/docs/api-reference/flows/video/create#request.body.veo-3.1-fast-generate-001) #### Google Veo 3.1 Lite A cost-efficient, high-speed variant of Veo 3.1 optimized for rapid generation at reduced compute. **Generation inputs:** * Text-to-Video * Start Frame **Features:** * Granular creative control with negative prompts and dedicated sound control * Fixed durations: 4s, 6s, and 8s * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 720p * Aspect ratios: 16:9, 9:16 **Ideal for:** * High-volume content creation where speed and cost efficiency are prioritized * Rapid concept exploration and previews **Cost:** Varies based on selected settings and duration #### Gemini Omni Flash 1.1 Google's physics-aware video model with native audio, up to 4K output, and both frame interpolation and reference-guided generation. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 10) * Video References (up to 3, each up to 10s) **Features:** * Frames and references are mutually exclusive: interpolate between frames, or guide with references * Fixed durations from 3s to 10s; duration follows the reference video when one is attached * Strong rendering of short on-screen text and labels * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 360p, 720p, 1080p, 4K * Aspect ratios: 16:9, 9:16 **Ideal for:** * Explainers and kinetic typography with legible text * Physically grounded motion with consistent subjects across references **Cost:** Varies based on selected settings and duration #### Gemini Omni Flash 1.1 Extend Continues an existing video from where it ends using Gemini Omni Flash 1.1, for a combined length of up to 40 seconds. **Generation inputs:** * Video to extend (required, up to 30s) * Text prompt describing the continuation **Features:** * Extension length from 3s to 10s * Aspect ratio follows the input video * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 360p, 720p, 1080p, 4K **Ideal for:** * Lengthening Gemini Omni Flash generations without a visible cut * Building sequences longer than a single generation allows **Cost:** Varies based on selected settings and duration #### Gemini Omni Flash The first Gemini Omni video model: physics-aware motion, native audio, and the strongest on-screen text rendering of the available video models, at 720p. **Generation inputs:** * Text-to-Video * Image References (up to 8) * Video Reference (1, up to 10s) **Features:** * Fixed durations from 3s to 10s; duration follows the reference video when one is attached * No start or end frame; use image references to anchor a subject instead * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 720p * Aspect ratios: 16:9, 9:16 **Ideal for:** * Short captions, titles, and labeled explainers * Multi-image consistency at 720p **Cost:** Varies based on selected settings and duration #### Seedance 1.5 Pro An upgraded specialized model for creating dynamic, high-fidelity sequences with enhanced temporal stability and precise transition control between keyframes. **Generation inputs:** * Text-to-Video * Start Frame * End Frame **Features:** * Seamlessly bridges start and end frames for coherent, multi-shot sequences * High-fidelity modeling of complex actions and environmental consistency * Supports fixed generation lengths from 4s up to 12s * Generate up to 4 variations at a time * Native support for enabled or disabled audio per generation **Output options:** * Resolutions: 420p, 720p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Storytelling and action scenes requiring stable physics between specific visual benchmarks * Cinematic transitions and professional video assets with strict start/end requirements **Cost:** Varies based on selected settings and duration > **Warning** > > Seedance 1.5 Pro is deprecated. ByteDance retires the model on November 11, 2026, and it is > no longer offered for new generations. Use a Seedance 2.0 model instead. Seedance 1.5 Pro is > not available in the United States. #### FLUX 3 Black Forest Labs' video model with natural motion, multi-shot scenes, generated audio, and video extension. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Video to extend (1, up to 15s) **Features:** * Video extension is mutually exclusive with start and end frames * Flexible generation lengths from 5s up to 20s * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 720p, 1080p * Aspect ratios: 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Multi-shot scenes from a single prompt * Extending an existing clip with matching motion and audio **Cost:** Varies based on selected settings and duration #### MiniMax H3 MiniMax's flagship video model with multimodal references, generated audio, and output up to 4K. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 9) * Video References (up to 3, each 2s to 15s, 15s combined) * Audio References (up to 3, each 2s to 15s, 15s combined) **Features:** * Combined cap of 12 references per generation; audio references require at least one image or video reference * Frames and references are mutually exclusive * Flexible generation lengths from 5s up to 15s * Generates synchronized audio * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 768p, 2K, 4K (2K and 4K are upscaled from 768p) * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Reference-driven scenes that need high-resolution delivery * Dialogue clips driven by reference audio **Cost:** Varies based on selected settings and duration > **Warning** > > MiniMax H3 is not available in the United States. #### MiniMax H3 Max A near-instant variant of MiniMax H3 with the same inputs and strong aesthetics, rendered natively at up to 768p. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 9) * Video References (up to 3, each 2s to 15s, 15s combined) * Audio References (up to 3, each 2s to 15s, 15s combined) **Features:** * Combined cap of 12 references per generation; audio references require at least one image or video reference * Frames and references are mutually exclusive * Flexible generation lengths from 5s up to 15s * Generates synchronized audio * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 768p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Rapid iteration on prompts and references * Previews before rendering with MiniMax H3 **Cost:** Varies based on selected settings and duration > **Warning** > > MiniMax H3 Max is not available in the United States. #### Kling O1 A state-of-the-art reasoning video model designed for superior prompt adherence and complex physical world simulation, utilizing advanced logical processing to interpret and execute intricate instructions. **Generation inputs:** * Text-to-Video (Description) * Start Frame & End Frame * Video Reference * Image Reference **Features:** * Exceptional ability to interpret multi-layered prompts and execute complex chronological actions * Leverages both images and videos as visual anchors to maintain high character and scene consistency * Superior modeling of physical interactions, cause-and-effect, and fluid dynamics * Supports high-quality generation for 5s and 10s clips * Generate up to 4 variations at a time **Output options:** * Resolutions: Dependant on the input * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * Highly specific creative concepts requiring precise adherence to long, detailed descriptions * Professional storytelling where physical realism and multi-reference consistency are critical **Cost:** Varies based on selected settings and duration > **Warning** > > Kling O1 is not available in the United States. #### Kling O1 Edit A natural language-driven video-to-video editing model that enables complex visual transformations—such as character replacement and environment swaps—without the need for manual masking or frame-by-frame adjustments. **Generation inputs:** * Source Video * Image References (up to 4 distinct elements/angles) * Text Description (Natural language instructions) **Features:** * Interprets conversational prompts to replace subjects or settings while respecting the original motion structure * Maintains original camera angles, movement patterns, and spatial relationships throughout the edit * Combines up to 4 total elements (including frontal and multi-angle images) to ensure high-fidelity character consistency * Option to preserve original source audio or generate silent output per generation * Generate up to 4 edited variations at a time **Output options:** * Resolutions: Dependent on source video * Aspect ratios: Matches source video **Ideal for:** * High-fidelity character replacement in existing footage while keeping original movements * Complete scene environment transformations (e.g., changing a daytime city to a futuristic nightscape) * Applying style transfers that require strict adherence to existing camera dynamics **Cost:** Varies based on video duration and selected settings > **Warning** > > Kling O1 Edit is not available in the United States. #### Kling 2.6 Motion Control A specialized model for precise motion transfer, allowing you to drive a character image with a reference video to replicate specific movements, gestures, and camera angles. **Generation inputs:** * Character Image (Source) * Motion Video (Reference) * Text Description (Optional) **Features:** * Choose "Match Video" for exact motion replication or "Match Image" for adding new creative motion to a character * Supports up to 30s in Match Video mode and 10s in Match Image mode * High-fidelity mapping of human movement from reference footage to a still character * Native support for enabling or disabling audio per generation * Generate up to 4 variations at a time **Output options:** * Resolutions: Dependent on the source * Aspect ratios: Dependent on the source **Ideal for:** * Replicating complex choreography or specific movements on a custom character * Long-form character animation requiring high motion fidelity * Social media content driven by trending video movements **Cost:** Varies based on selected settings and duration > **Warning** > > Kling 2.6 Motion Control is not available in the United States. #### Kling 3.0 Motion Control A specialized model for precise motion transfer built on the Kling 3.0 architecture, allowing you to drive a character image with a reference video to replicate specific movements, gestures, and camera angles with enhanced fidelity. **Generation inputs:** * Character Image (Source) * Motion Video (Reference) * Text Description (Optional) **Features:** * Choose "Match Video" for exact motion replication or "Match Image" for adding new creative motion to a character * Supports up to 30s in Match Video mode and 10s in Match Image mode * High-fidelity mapping of human movement from reference footage to a still character * Native support for enabling or disabling audio per generation * Generate up to 4 variations at a time **Output options:** * Resolutions: Dependent on the source * Aspect ratios: Dependent on the source **Ideal for:** * Replicating complex choreography or specific movements on a custom character * Long-form character animation requiring high motion fidelity * Social media content driven by trending video movements **Cost:** Varies based on selected settings and duration > **Warning** > > Kling 3.0 Motion Control is not available in the United States. #### Kling 2.6 An optimized generative model designed for enhanced motion fidelity and smoother transitions, providing a balance between high-speed iteration and production-quality visual output. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Enhanced motion dynamics: Significant improvements in movement fluidity and realistic physics interactions * Flexible sound control: Native support for enabling or disabling audio per generation * Batch creation: Generate up to 4 variations simultaneously * Granular refinement: Advanced creative control through negative prompt support * Fixed durations: Supports generation lengths of 5s and 10s **Output options:** * Resolutions: Dependant on the input * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * High-action clips requiring fluid character movement * Professional-grade social media content with strong prompt adherence **Cost:** Varies based on selected settings and duration > **Warning** > > Kling 2.6 is not available in the United States. #### Kling 2.5 A balanced and versatile model for high-quality, full-HD video generation. **Generation inputs:** * Text-to-Video * Start Frame **Features:** * Excels at simulating complex motion and realistic physics * Accurately models fluid dynamics and expressions * Fixed durations: 5s and 10s * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1080p * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * Realistic physics simulations and complex motion **Cost:** Varies based on selected settings and duration > **Note** > > End frame is not currently supported. Cannot provide image references. Sound control not available. > **Warning** > > Kling 2.5 is not available in the United States. #### Runway Gen-4.5 State-of-the-art motion quality, prompt adherence, and visual fidelity for cinematic, highly realistic video. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Exceptional motion quality with industry-leading realism and physics simulation * Superior prompt adherence for precise creative control over complex scenes * High visual fidelity delivering cinematic-grade output * Generate up to 4 variations simultaneously **Output options:** * Resolutions: 720p * Aspect ratios: 16:9, 9:16 **Ideal for:** * Cinematic content requiring the highest motion quality and realism * Professional productions demanding precise prompt adherence * High-fidelity visual storytelling **Cost:** Varies based on selected settings and duration #### Runway Gen-4 Turbo An advanced video model designed for rapid iteration and cost-effective creation, capable of producing high-quality videos. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Optimized for ultra-fast generation, delivering results up to four times faster than previous iterations * Allows for precise direction of character movements, camera angles, and scene compositions * Excels at maintaining visual coherence and stability across dynamic scenes * Supports generation lengths ranging from 2s up to 10s * Generate up to 4 variations simultaneously **Output options:** * Resolutions: 720p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Rapid prototyping and creative experimentation requiring near-instant feedback * Professional projects needing quick turnarounds for high-resolution marketing assets * Cinematic content with specific camera movement requirements **Cost:** Varies based on selected settings and duration #### Runway Gen-4 Aleph A state-of-the-art in-context video model designed for multi-task visual generation, capable of performing complex edits while maintaining the underlying structure of the source footage. **Generation inputs:** * Source Video * Reference Image * Text Description **Features:** * Seamlessly add, remove, or transform objects and subjects within a scene with natural lighting, shadows, and perspective * Change locations, seasons, and time of day (e.g., converting cloudy footage to a dramatic sunset) with realistic color temperature updates * Modify the age and appearance of actors or retexture clothing and subjects through simple natural language prompts * Apply the specific motion and camera path of a reference video to a static image for precise animation control * Generate entirely new camera angles, such as reverse shots or low angles, from a single existing video sequence * Includes precise green-screening (isolation with edge detection), next-shot generation for story continuation, and aesthetic style transfer * Generate up to 4 variations simultaneously **Output options:** * Resolutions: 720p * Aspect ratios: Auto **Ideal for:** * Professional visual effects tasks like digital de-aging, relighting, and object removal * Rapid cinematic prototyping and generating alternative camera coverage from a single shot * Creative marketing content requiring drastic environmental or stylistic transformations **Cost:** Varies based on selected settings and duration #### Runway Aleph 2.0 Runway's video editing model: transforms an existing video toward a target look while preserving motion and consistency. **Inputs:** * Source video (required, 2s to 30s) * Target look image (required) * Text prompt describing the edits **Features:** * Output length and framing follow the source video * Style transfer guided by the target look image * Batch creation with up to 4 generations at a time **Ideal for:** * Relighting, restyling, or changing the setting of existing footage * Applying a reference image's look to a whole clip **Cost:** Varies based on selected settings and duration #### Runway Act-Two A specialized performance-transfer model that animates characters by mapping the motion, speech, and facial expressions from a driving video onto a character image or video reference. **Generation inputs:** * Driving Performance (Video) * Character Input (Image or Video) **Features:** * Transfers nuanced facial expressions, lip-sync, and synchronized audio directly from a source actor to any character * Automatically adds secondary motion and subtle camera shakes to static character images for a more natural look * Precise toggle to enable or disable body and hand movements when using a character image * Adjustable settings to balance between intense emotional performance and character visual consistency * Ability to change the character's voice after generation while maintaining perfect alignment with the driving performance **Output options:** * Resolutions: 720p * Aspect ratios: Auto **Ideal for:** * Bringing static character portraits to life with realistic human motion and speech * Animating non-human characters or stylized avatars with high-fidelity expressions * Rapidly producing talking-head content with integrated body gestures **Cost:** Varies based on selected settings and duration #### Seedance 1 Pro A specialized model for creating dynamic, multi-shot sequences with large movement and action. **Generation inputs:** * Text-to-Video * Start Frame * End Frame **Features:** * Highly stable physics and seamless transitions between shots * Fixed durations: 3s, 4s, 5s, 6s, 7s, 8s, 9s, 10s, 11s, and 12s * Batch creation with up to 4 generations at a time * Maximum creative flexibility with numerous aspect ratio options **Output options:** * Resolutions: 480p, 720p, 1080p * Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Storytelling and action scenes requiring stable physics **Cost:** Varies based on selected settings and duration > **Note** > > Aspect ratio and resolution do not affect generation credits, but duration does. > **Warning** > > Seedance 1 Pro is not available in the United States. #### LTX Audio-to-Video A DiT-based foundation model designed to generate synchronized video and audio in a single pass, ensuring coherent speech and realistic motion. **Generation inputs:** * Text-to-Video * Image-to-Video * Audio-to-Video * Depth-to-Video **Features:** * Generates dialogue, lip movement, and ambient audio simultaneously for perfect alignment without external tools * Dynamic scenes with stable motion, consistent identity, and strong frame-to-frame coherence * Supports high-fidelity synchronized generation for up to 20 seconds * Advanced creative direction through granular negative prompt support * Generate up to 4 variations at a time **Output options:** * Resolutions: 720p, 1080p * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Coherent speech and expressive character performances * Narrative content requiring integrated ambient audio and consistent timing * Dynamic scenes with complex camera-aware motion logic **Cost:** Varies based on selected settings and duration #### LTX 2 Pro A high-fidelity generative model optimized for maximum visual detail and structural stability, capable of producing production-grade 4K output with fluid motion. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Prioritizes visual quality and consistency over speed, ensuring stable results across extended sequences * Supports both 25 FPS and 50 FPS for exceptionally smooth and professional motion * Integrated audio-visual generation with a toggle for sound on or off * Built to handle native 1080p, 2k, and 4k outputs without loss of detail * Generate up to 4 variations at a time **Output options:** * Resolutions: 1080p, 2k, 4k * Frame rates: 25 FPS, 50 FPS * Aspect ratio: 16:9 (Default) * Durations: 6s, 8s, 10s **Ideal for:** * High-resolution cinematic production requiring 4K clarity * Professional content necessitating smooth 50 FPS motion * Detailed sequences where visual stability and structural integrity are critical **Cost:** Varies based on selected settings and duration #### LTX 2 Retake A precision AI directing tool that allows for targeted redirection of dialogue, emotion, and action within existing shots without breaking continuity or regenerating the entire sequence. **Generation inputs:** * Source Video * Text Description **Features:** * Modify specific segments while maintaining strong context preservation from surrounding frames * Rephrase spoken lines while keeping the character's voice, performance, and environment consistent * Multiple edit modes: Select between "Audio & Video," "Audio only," or "Video only" to isolate and regenerate specific elements of the shot * New content naturally inherits the original motion, lighting, and tone for seamless transitions * Instantly experiment with alternate character reactions, emotional beats, or camera movements within a single shot * Generate up to 4 variations simultaneously for side-by-side creative comparison **Output options:** * Aspect ratios: 16:9 only **Ideal for:** * Adjusting scripts and refining dialogue without the need for reshoots or rerecording * Fixing emotional beats or pacing issues in post-production * Testing multiple brand messages and calls-to-action within a single marketing asset **Cost:** Varies based on selected settings and duration #### LTX 2 Fast A speed-optimized generative model built for tight feedback loops and high-velocity content creation, delivering high-resolution visuals with significantly reduced render times. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Engineered for speed and rapid iteration, allowing for quick visual experimentation and near-instant previews * Supports native 1080p, 2k, and 4k outputs with lower compute overhead than the Pro model * Capabilities for both 25 FPS and 50 FPS for smooth motion at high speeds * Enables rapid generation of synchronized audio-visual content for durations up to 20 seconds * Native support for enabling or disabling audio per generation * Generate up to 4 variations simultaneously **Output options:** * Resolutions: 1080p, 2k, 4k * Frame rates: 25 FPS, 50 FPS * Aspect ratio: 16:9 (Default) * Durations: 6s, 8s, 10s, 12s, 14s, 16s, 18s, 20s **Ideal for:** * Rapid prototyping and creative exploration where speed is prioritized over maximum detail * High-volume social media content requiring quick turnarounds * A/B testing different visual concepts and motion styles **Cost:** Varies based on selected settings and duration #### Wan 3.0 A cinematic video model with image, video, and audio references, generated audio, and generations up to 30 seconds. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 10) * Video References (up to 5, 15s combined) * Audio References (up to 5, 15s combined) **Features:** * Frames and references are mutually exclusive * Fixed durations: 5s, 10s, 15s, 20s, and 30s * Native sound control with audio enabled or disabled per generation * Optional prompt enhancement * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p, 1080p * Aspect ratios: Auto, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Longer cinematic shots with strong prompt adherence * Reference-driven scenes with audio **Cost:** Varies based on selected settings and duration > **Warning** > > Wan 3.0 is not available in the United States. #### Wan 3.0 Prime A faster variant of Wan 3.0 with the same inputs and output options, suited to ideation. **Generation inputs:** * Text-to-Video * Start Frame * End Frame * Image References (up to 10) * Video References (up to 5, 15s combined) * Audio References (up to 5, 15s combined) **Features:** * Roughly half the generation time of Wan 3.0 * Fixed durations: 5s, 10s, 15s, 20s, and 30s * Native sound control with audio enabled or disabled per generation * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p, 1080p * Aspect ratios: Auto, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Exploring ideas before a final Wan 3.0 render * Turnaround-sensitive workflows **Cost:** Varies based on selected settings and duration > **Warning** > > Wan 3.0 Prime is not available in the United States. #### Wan 2.6 A next-generation cinematic video platform that utilizes a unified multimodal architecture to deliver production-ready 1080p content with native audio synchronization and intelligent multi-shot sequencing. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) * Video Reference (Video-to-Video) * Audio Reference (Optional background audio or dialogue) **Features:** * Unified multimodal system: Processes text, images, video, and audio through a single integrated framework for consistent output quality * Native audio sync: Automatically generates and aligns dialogue, narration, and environmental sound effects with on-screen movement * Intelligent multi-shot sequencing: Automatically organizes connected video sequences into coherent story arcs while maintaining character consistency * Extended durations: Supports stable, high-quality generation for fixed lengths of 5s, 10s, and 15s * Advanced creative control: Supports negative prompts for granular detail management and batch creation of up to 4 variations **Output options:** * Resolutions: 720p, 1080p * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Professional narrative storytelling and complex multi-shot cinematic sequences * Social media ads and marketing content requiring integrated, high-fidelity audio * Character-driven content requiring strict visual and motion consistency via video references **Cost:** Varies based on selected settings and duration > **Warning** > > Wan 2.6 is not available in the United States. #### Wan 2.5 Video A versatile model that delivers cinematic motion and high prompt fidelity from text or a starting image. **Generation inputs:** * Text-to-Video * Start Frame (Image-to-Video) **Features:** * Granular creative control with negative prompts and dedicated sound control * Fixed durations: 5s and 10s * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 480p, 720p, 1080p * Aspect ratios: 16:9, 1:1, 9:16 **Ideal for:** * Cinematic content with strong prompt adherence **Cost:** Varies based on selected settings and duration > **Note** > > Generation cost varies based on selected settings. > **Warning** > > Wan 2.5 Video is not available in the United States. #### Image generative models #### GPT Image 2.5 Sunburst OpenAI's production-grade image model for highly detailed output and precise editing. **Generation inputs:** * Text-to-Image * Image References (up to 10) **Features:** * Five quality levels from Low to Maximum * Custom resolution up to 3840px per side, at aspect ratios up to 3:1 * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K, 2K, 4K, or custom * Aspect ratios: 3:1, 21:9, 2:1, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16, 1:2, 1:3 **Ideal for:** * Final assets where detail and fidelity matter most * Precise edits to existing images **Cost:** Varies based on selected settings and number of variations **API:** [`gpt-image-2.5-sunburst`](/docs/api-reference/flows/image/create#request.body.gpt-image-2.5-sunburst) #### GPT Image 2.5 Flare A fast GPT Image 2.5 variant with the same controls as Sunburst, tuned for everyday, high-volume generation. **Generation inputs:** * Text-to-Image * Image References (up to 10) **Features:** * Five quality levels from Low to Maximum * Custom resolution up to 3840px per side, at aspect ratios up to 3:1 * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K, 2K, 4K, or custom * Aspect ratios: 3:1, 21:9, 2:1, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16, 1:2, 1:3 **Ideal for:** * High-volume image generation * Quick iterations that still need 4K and custom sizes **Cost:** Varies based on selected settings and number of variations **API:** [`gpt-image-2.5-flare`](/docs/api-reference/flows/image/create#request.body.gpt-image-2.5-flare) #### GPT Image 2 An advanced AI model designed for precise image generation with enhanced text rendering and high-resolution output capabilities. **Features:** * Precise text control for creating images with accurate typography and clarity * High-resolution PNG output suitable for professional and commercial use * Supports multiple image references to guide style and composition * Generates up to 4 images at a time **Output options:** * Aspect ratios: 3:2, 1:1, 2:3 * Quality options: low, medium, high **Ideal for:** * Marketing visuals and design assets requiring precise text placement * Professional content with high-resolution requirements * Creative projects needing accurate text-based image control **Cost:** Varies based on selected settings and number of variations **API:** [`gpt-image-2`](/docs/api-reference/flows/image/create#request.body.gpt-image-2) #### Nano Banana 2 An advanced AI image generation model combining production-ready quality with ultra-fast processing, featuring improved world knowledge and subject consistency. **Features:** * Native 4K resolution output without upscaling for crystal-clear detail * Lightning-fast image generation in as little as 1-2 seconds * Superior prompt adherence through advanced reasoning and instruction following * Crisp, accurate text rendering in multiple languages for mockups, signage, and infographics * Consistent multi-subject styling that preserves identity across multiple characters and objects * Improved world knowledge for more accurate scene generation * Supports multiple image references to guide generation * Generates up to 4 images at a time **Output options:** * Aspect ratios: 21:9, 16:9, 5:4, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16 * Resolutions: Up to 4K **Ideal for:** * Rapid creative workflows requiring real-time iteration * Brand content and storytelling with consistent character identity * Professional visuals with accurate text and typography **Cost:** Varies based on selected settings and number of variations **API:** [`gemini-3.1-flash-image`](/docs/api-reference/flows/image/create#request.body.gemini-3.1-flash-image) #### Nano Banana 2 Lite A fast, low-cost variant of Nano Banana 2 for rapid generation and editing at 1K. **Generation inputs:** * Text-to-Image * Image References (up to 14) **Features:** * Fixed 1K output for consistent speed and cost * Same 14-image reference capacity as Nano Banana 2 * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K * Aspect ratios: 21:9, 16:9, 5:4, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16 **Ideal for:** * Rapid iteration and drafts * Bulk edits where 1K is sufficient **Cost:** Varies based on selected settings and number of variations **API:** [`gemini-3.1-flash-lite-image`](/docs/api-reference/flows/image/create#request.body.gemini-3.1-flash-lite-image) #### Krea 2 Medium A versatile, production-ready image generation model from Krea AI, balancing quality and speed for a wide range of creative workflows. **Features:** * High-fidelity image generation with strong prompt adherence * Supports multiple image references to guide generation * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Professional content creation requiring consistent quality * Rapid iteration across diverse creative concepts **Cost:** Varies based on selected settings and number of variations #### Krea 2 Large Krea AI's flagship image generation model, delivering maximum visual fidelity and advanced creative control for professional production workflows. **Features:** * Exceptional detail and realism for professional-grade output * Advanced style control with support for multiple image references * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * High-end commercial and editorial image production * Complex creative compositions requiring maximum fidelity **Cost:** Varies based on selected settings and number of variations #### Recraft V4.1 A design-focused text-to-image model with strong prompt control and clean composition. **Generation inputs:** * Text-to-Image **Features:** * 2K output uses the Pro model variant * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K, 2K * Aspect ratios: 2:1, 16:9, 3:2, 14:10, 4:3, 5:4, 1:1, 4:5, 3:4, 10:14, 2:3, 6:10, 9:16, 1:2 **Ideal for:** * Layout-driven graphics, icons, and posters * Clean compositions from text alone **Cost:** Varies based on selected settings and number of variations #### Recraft V4 A design-focused image model that can match the style of reference images. **Generation inputs:** * Text-to-Image * Style References (up to 10) **Features:** * Style match control: Precise follows the reference style closely, Flexible matches the general look * 2K output uses the Pro model variant * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K, 2K * Aspect ratios: 2:1, 16:9, 3:2, 14:10, 4:3, 5:4, 1:1, 4:5, 3:4, 10:14, 2:3, 6:10, 9:16, 1:2 **Ideal for:** * Brand-consistent graphics from a set of style references * Illustration and design work **Cost:** Varies based on selected settings and number of variations #### Seedream 5 Lite A lightweight, fast-generation image model from ByteDance delivering high-quality results with reduced compute requirements. **Features:** * Fast generation for rapid creative iteration * Supports multiple image references to guide generation * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * High-volume image creation workflows requiring speed * Rapid concept exploration and previews **Cost:** Varies based on selected settings and number of variations **API:** [`bytedance-seedream-5-lite`](/docs/api-reference/flows/image/create#request.body.bytedance-seedream-5-lite) > **Warning** > > Seedream 5 Lite is not available in the United States. #### Seedream 5 Pro The higher-fidelity Seedream 5 variant for precise editing, multilingual text, and dense infographics. **Generation inputs:** * Text-to-Image * Image References (up to 10) **Features:** * Accurate multilingual text rendering * Handles dense layouts such as infographics and posters * Batch creation with up to 4 generations at a time **Output options:** * Resolutions: 1K, 2K * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Text-heavy designs in multiple languages * Multi-image editing with strict reference adherence **Cost:** Varies based on selected settings and number of variations **API:** [`bytedance-seedream-5-pro`](/docs/api-reference/flows/image/create#request.body.bytedance-seedream-5-pro) > **Warning** > > Seedream 5 Pro is not available in the United States. #### GPT Image 1.5 A high-speed flagship model designed for precise text-based image generation and complex, non-destructive photo editing that preserves original details. **Features:** * Reliably executes requested changes while maintaining the integrity of lighting, composition, and subject appearance within source images * Supports complex editing tasks including adding, subtracting, combining, and blending elements * Delivers outputs up to 4x faster than previous iterations * Generates up to 4 images at a time **Output options:** * Aspect ratios: 3:2, 1:1, 2:3 * Quality options: low, medium, high **Ideal for:** * Practical photo adjustments and realistic virtual try-ons for clothing or hairstyles * Conceptual transformations and stylistic filters that retain the essence of the input image * Rapid iteration of text-to-image concepts **Cost:** Varies based on selected settings and number of variations **API:** [`gpt-image-1.5`](/docs/api-reference/flows/image/create#request.body.gpt-image-1.5) #### GPT Image 1 OpenAI's first-generation image model with good prompt adherence, legible text, and detailed editing. **Generation inputs:** * Text-to-Image * Image References (up to 5) **Features:** * Three quality levels: Low, Medium, High * Batch creation with up to 4 generations at a time **Output options:** * Aspect ratios: 3:2, 1:1, 2:3 **Ideal for:** * Workflows already built around GPT Image 1 output * Simple edits and text-in-image tasks at standard resolution **Cost:** Varies based on selected settings and number of variations **API:** [`gpt-image-1`](/docs/api-reference/flows/image/create#request.body.gpt-image-1) #### Seedream 4.5 A high-performance multimodal foundation model that unifies text-to-image synthesis, precise image editing, and complex multi-image composition into a single, efficient framework. **Features:** * Native support for fast generation of high-fidelity images up to 4K resolution * Exceptional preservation of facial features, lighting, color tone, and fine details during editing tasks based on reference inputs * Accurately identifies and blends target elements across multiple input images for controllable, consistent results * Offers designer-level composition capabilities with clear, accurate rendering of small text for posters and brand visuals * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 * Resolutions: 2K, 4K **Ideal for:** * Professional graphic design workflows requiring precise layout and typography * Complex photo editing needing strict adherence to reference identity and lighting * High-resolution creative compositing using multiple visual sources **Cost:** Varies based on selected settings and number of variations > **Warning** > > Seedream 4.5 is not available in the United States. #### Kling O1 Image A high-fidelity image generation model with advanced reasoning capabilities, designed for superior prompt adherence and precise visual consistency across complex compositions. **Features:** * Exceptional ability to interpret and execute intricate, multi-layered text descriptions with high accuracy * Leverages image references to maintain subject identity, lighting, and aesthetic style across generations * Optimized for realistic textures, complex spatial relationships, and professional-grade lighting effects * Generates up to 4 images at a time **Output options:** * Aspect ratios: Auto, 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9 * Resolutions: 1K, 2K **Ideal for:** * Professional creative assets requiring strict adherence to detailed and technical text prompts * High-consistency image editing and character design using visual references **Cost:** Varies based on selected settings and number of variations > **Warning** > > Kling O1 Image is not available in the United States. #### FLUX.2 \[Pro] A production-grade image generation and editing model designed for professional workflows, offering state-of-the-art visual quality with a focus on speed, precision, and consistency. **Features:** * Reference multiple images simultaneously to achieve industry-leading character and identity consistency across hundreds of assets * Provides an unprecedented leap in detail quality, closing the gap with real photography for everything from fabric textures to architectural elements * Delivers production-ready text rendering for complex typography, UI mockups, and infographics * Supports precise brand color specification via hex codes with no approximation * Ensures accurate object positioning, realistic physics, coherent lighting, and proper perspective throughout complex scenes * Optimized for better accuracy and responsiveness to structured, complex instructions * Generates up to 4 images at a time **Output options:** * Aspect Ratios: 16:9, 4:3, 1:1, 3:4, 9:16 * Resolutions: 720p, 1080p, 2K **Ideal for:** * Running character-consistent campaigns and placing products accurately in any context * Creating interface mockups with readable text and consistent visual design systems * Generating product photography at scale and contextual lifestyle shots **Cost:** Varies based on selected settings and number of variations #### Nano Banana Pro A professional-grade, reasoning-based image generation and editing model designed for high-fidelity asset production, advanced creative control, and precise instruction following. **Inputs:** * Image Reference * Text Description (supports complex, multi-layered prompts) **Features:** * Plans scenes before rendering to deliver physics-accurate lighting, accurate object relationships, and superior prompt adherence * Generates sharp, legible multilingual text in various font styles and handwriting for impactful posters and product mockups * Maintains high fidelity and resemblance for up to 5 people and multiple objects across diverse creative outputs * Integrates Google Search to enhance visuals with actual data, real-world knowledge, and real-time information like weather or sports * Adjust camera angles, focal points, and scene lighting (e.g., transforming day to night) with advanced localized editing tools * Superior spatial understanding enables the generation of accurate infographics, technical diagrams, and presentation slides * Generates up to 4 variations at a time **Output options:** * Resolutions: 1K, 2K, 4K * Aspect ratios: Auto, 21:9, 16:9, 5:4, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16 **Ideal for:** * Professional advertising, brand assets, and high-end e-commerce product photography * Educational explainers, data-driven infographics, and complex technical documentation * Rapid prototyping of high-resolution visual designs with consistent character or brand identity **Cost:** Varies based on selected settings and number of variations **API:** [`gemini-3-pro-image`](/docs/api-reference/flows/image/create#request.body.gemini-3-pro-image) #### Nano Banana A high-speed model for quick, high-quality image generation and editing directly from text prompts. **Features:** * Supports multiple image references to guide generation * Generates up to 4 images at a time **Output options:** * Aspect ratios: 21:9, 16:9, 5:4, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16 **Ideal for:** * Rapid image creation and iteration **Cost:** Varies based on selected settings and number of variations **API:** [`gemini-2.5-flash-image`](/docs/api-reference/flows/image/create#request.body.gemini-2.5-flash-image) #### Runway Gen-4 Image An advanced base model designed for high-fidelity image generation, offering unprecedented stylistic control and visual memory to maintain consistency across scenes. **Features:** * Anchor characters, styles, or specific objects using input images to maintain professional-grade consistency across multiple outputs * Optimized to interpret complex, natural language descriptions for precise control over visual details, lighting, and emotions * Capable of generating high-quality visuals for diverse use cases, from cinematic storyboards to professional product photography * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 * Resolutions: 720p, 1080p **Ideal for:** * Ensuring protagonists retain their appearance across different environments and lighting treatments * Rapidly producing high-resolution brand assets, virtual try-ons, and scale-ready product visualizations * Exploring varied artistic directions while locking in core visual identity via image references **Cost:** Varies based on selected settings and number of variations #### Runway Gen-4 Image Turbo An optimized image generation model engineered for speed, delivering results 2.5x faster than the standard Gen-4 Image while maintaining identical output quality. **Features:** * Optimized processing allows for rapid creative exploration, generating high-fidelity images in a fraction of the time * Upload up to three reference images to guide the model's understanding of specific characters, environments, and artistic styles * Solves the challenge of visual drift by encoding specific visual characteristics from reference images to maintain identity across multiple generations * Effortlessly apply the aesthetic, lighting, and texture of a reference image to entirely new subjects and scenes * Utilize seed parameters to systematically explore variations or recreate specific outputs with precision * Generates up to 4 images at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 * Resolutions: 720p, 1080p **Ideal for:** * Rapidly producing platform-optimized visuals for Instagram Stories (9:16) or standard posts (1:1) at scale * Maintaining strict visual identity across campaigns by using brand style guides as reference images * Developing consistent character poses and settings while ensuring the protagonist remains recognizable * Creating diverse lifestyle and seasonal product shots accurately through reference-based guidance **Cost:** Varies based on selected settings and number of variations #### Seedream 4 A specialized image model for generating multi-shot sequences or scenes with large movement and action. **Features:** * Excels at creating images with stable physics and coherent transitions * Supports multiple image references to guide generation * Generates up to 4 images at a time **Output options:** * Aspect ratios: auto, 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Action scenes and dynamic compositions **Cost:** Varies based on selected settings and number of variations > **Warning** > > Seedream 4 is not available in the United States. #### Wan 2.5 Image The image variant of Wan 2.5, with strong prompt fidelity and support for reference images. **Generation inputs:** * Text-to-Image * Image References (up to 2) **Features:** * Negative prompt and seed controls * Batch creation with up to 4 generations at a time **Output options:** * Aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16 **Ideal for:** * Stills that match the look of Wan video generations * Prompt-faithful concept images **Cost:** Varies based on selected settings and number of variations > **Warning** > > Wan 2.5 Image is not available in the United States. #### FLUX.1 Kontext \[Pro] A professional model for advanced image generation and editing, offering strong scene coherence and style control. **Features:** * Image-based style control requiring a reference image to guide visual aesthetic * Generates up to 4 images at a time **Output options:** * Aspect ratios: 21:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 4:5, 9:16, 9:21 **Ideal for:** * Professional content with precise style requirements **Cost:** Varies based on selected settings and number of variations #### Lip-sync models #### Creatify Aurora A state-of-the-art diffusion transformer (DiT) model designed for rendering ultra-realistic, reactive avatars driven by audio and text guidance. **Generation inputs:** * Avatar (Source Image) * Speech (Audio File) * Text Description (Guidance) **Features:** * Goes beyond basic lip-sync to include context-aware blinking, breathing, and natural facial expressions * Automatically synchronizes hand and full-body movements based on vocal tone and inflection for a studio-grade performance * Accurately interprets vocal intensity and pitch to deliver performance-accurate emotional expressions * Maintains high character fidelity and behavioral coherence even across extended dialogue or musical performances * Optimized for various setups, including side-angle presentations, podcast-style dialogues, and stylized animations * Generate up to 4 variations at a time **Output options:** * Resolutions: 480p, 720p **Ideal for:** * Professional avatar-based video ads and marketing content * High-fidelity virtual storytelling and expressive musical performances * Long-form educational or training videos requiring consistent character presence **Cost:** Varies based on input, settings and duration **API:** [`creatify-aurora`](/docs/api-reference/flows/video/create#request.body.creatify-aurora) #### Veed Fabric A fast, precise image-to-video lip-sync model that animates a still image to match provided audio. **Inputs:** * Source image * Speech audio file **Features:** * Animates the mouth and facial expressions on the source image to match provided audio * Creates high-fidelity "talking" video from still images * Lip-sync specific tool, not a full video generation model **Ideal for:** * Creating talking avatars from still images * Adding dialogue to character portraits * Professional avatar-driven content workflows **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the image should contain a detectable figure. #### HeyGen Avatar 4 HeyGen's latest avatar model for generating highly realistic, expressive talking-head videos from a single image and audio input. **Inputs:** * Source image * Speech audio file **Features:** * Animates facial expressions and lip movements to match provided audio with high fidelity * Produces natural head motion and subtle micro-expressions for lifelike results * Lip-sync specific tool, not a full video generation model **Ideal for:** * Professional spokesperson and presenter videos * Creating personalized avatar content at scale * Marketing and training videos with consistent character presence **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the image should contain a clearly visible face. #### Sync 3 The latest generation video lip-sync model from Sync, delivering studio-quality synchronization across a wide range of content types. **Generation inputs:** * Source Video * Speech Audio **Features:** * High-accuracy lip-sync that preserves unique facial details, natural teeth, and skin texture * Optimized for all content types including live-action, 3D animation, and AI-generated video * Video-to-video lip-sync tool, not a full video generator **Ideal for:** * Professional dubbing and localization for film and advertising * Enhancing or correcting dialogue in any video format * High-volume content translation workflows **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the video should contain a detectable figure. #### Sync Lipsync 2 Pro A state-of-the-art video editing model designed for studio-grade lip-syncing that preserves unique facial details while scaling to high-resolution outputs. **Generation inputs:** * Source Video * Speech Audio **Features:** * Incorporates advanced upscaling to support 4K output while maintaining sharp, natural textures * Protects unique facial features such as natural teeth, freckles, makeup, and complex facial hair without loss of clarity * Optimized to work across all content types, including live-action, 3D animation, and AI-generated video * Delivers expressive, synchronized results immediately without requiring speaker-specific training or model fine-tuning * Generate up to 4 variations simultaneously **Ideal for:** * Professional-grade dubbing and localized content for film and high-end advertising * Enhancing or correcting dialogue in 3D animated and AI-generated characters * High-resolution projects requiring pixel-perfect facial consistency and detail **Cost:** Varies based on input, settings and duration #### OmniHuman 1.5 A dedicated utility model for generating exceptionally realistic, humanlike lip-sync. **Inputs:** * Static source image * Speech audio file **Features:** * Animates the mouth on the source image to match provided audio * Creates high-fidelity "talking" video from still images * Lip-sync specific tool, not a full video generation model **Ideal for:** * Creating talking avatars * Adding dialogue to still images * Professional dubbing workflows **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the image should contain a detectable figure. > **Warning** > > OmniHuman 1.5 is not available in the United States. #### Veed Lipsync A fast, affordable, and precise utility model for applying realistic lip-sync to videos. **Inputs:** * Source video * New speech audio file **Features:** * Re-animates mouth movements in source video to match new audio * Video-to-video lip-sync tool, not a full video generator **Ideal for:** * High-volume, cost-effective dubbing * Translating content * Correcting audio in video clips with realistic results **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the video should contain a detectable figure. #### Veed Lipsync 2.0 The higher-quality successor to Veed Lipsync for applying realistic lip-sync to existing videos. **Inputs:** * Source video * New speech audio file **Features:** * Re-animates mouth movements in the source video to match new audio * Video-to-video lip-sync tool, not a full video generator * Batch creation with up to 4 generations at a time **Ideal for:** * Dubbing and translation where output quality matters more than cost * Correcting dialogue in finished clips **Cost:** Varies based on input, settings and duration > **Note** > > For best results, the video should contain a detectable figure. #### Utility models #### Topaz Upscale A dedicated utility model for image and video upscaling, designed to enhance resolution and detail up to 4x. **Features:** * Enhancement tool that processes existing media * Increases media size while preserving natural textures and minimizing artifacts * Highly granular upscale factors: 1x, 1.25x, 1.5x, 1.75x, 2x, 3x, 4x * Video-specific: Flexible frame rate control (keep source or convert to 24, 25, 30, 48, 50, or 60 fps) **Ideal for:** * Improving quality of generated media * Restoring legacy footage or photos * Preparing assets for high-resolution displays **Cost:** Varies depending on input #### Background Removal Removes the background from an image and returns it with alpha transparency. **Inputs:** * Source image **Features:** * No prompt required * Single output per run **Ideal for:** * Product cutouts and compositing * Preparing subjects for use as references in other generations **Cost:** Varies depending on input #### Runway Ruby Converts standard dynamic range video into HDR. **Inputs:** * Source video (up to 30s) **Features:** * No prompt required * Single output per run; length and framing follow the source **Ideal for:** * Preparing footage for HDR displays * Re-grading generated video for delivery **Cost:** Varies based on video duration