Audio detector
Detect AI-generated audio with SynthID watermarking technology
Overview
ElevenLabs provides two tools for identifying AI-generated audio:
- Audio Detector: Uses Google DeepMind’s SynthID watermarking technology to detect imperceptible watermarks embedded in ElevenLabs-generated audio. Free for signed-in users. The tool first checks for a SynthID watermark, then falls back to the AI Speech Classifier if no watermark is found.
- AI Speech Classifier: A legacy tool that uses machine learning to analyze audio characteristics and predict whether content was generated by ElevenLabs. Available without sign-in.
The Audio Detector is the recommended tool for checking ElevenLabs-generated content. It provides more reliable detection through digital watermarking rather than statistical analysis.
What is SynthID?
SynthID is Google DeepMind’s digital watermarking technology designed to identify AI-generated content. It embeds imperceptible patterns into audio at frequencies masked by existing sound, making the watermark completely inaudible while maintaining audio quality.
SynthID watermarking is currently applied to all Text to Speech generations by free users and select paid features. The rollout is expanding in phases.
Using the audio detector
The Audio Detector can only detect SynthID watermarks embedded by ElevenLabs. It will not detect watermarks from other AI platforms. Additionally, content generated by third-party models available through ElevenLabs is generally not watermarked with SynthID.
Difference between audio detector and AI speech classifier
The Audio Detector is the recommended tool for signed-in users. It checks for a SynthID watermark first, then falls back to the AI Speech Classifier if no watermark is found. This tool returns a positive result if it detects either a SynthID watermark or receives a positive identification from the AI Speech Classifier.
The standalone AI Speech Classifier page remains available separately without requiring sign-in. This legacy tool uses machine learning to analyze audio characteristics and predict whether content was likely generated by ElevenLabs. However, this approach is statistical and may produce false positives or miss modified audio.
For signed-in users, the AI Speech Classifier’s functionality is now built into the Audio Detector, providing a more comprehensive detection method.
Watermarking details
Audio quality and latency
SynthID watermarking does not affect audio quality and adds zero latency in nearly all cases. The watermarks are imperceptible and carefully engineered to maintain the same generation speed.
Platform compatibility
No other platforms can currently detect ElevenLabs SynthID watermarks. The SynthID partnership is focused on public AI transparency and is not connected to other platforms’ monetization or content policies.
Third-party models
Content generated by third-party models available through ElevenLabs is generally not watermarked with SynthID unless explicitly stated otherwise. The integration is specific to audio generated directly by ElevenLabs models.
FAQ
Is the audio detector free to use?
Yes, the audio detector is free for all signed-in users.
Can I hear the SynthID watermark?
No. The watermark is completely inaudible. SynthID hides patterns at frequencies masked by existing audio, ensuring no perceptible difference from unwatermarked content.
Is my audio being watermarked right now?
SynthID watermarking is rolling out in phases. It currently covers all Text to Speech generations by free users and select paid features. If you are unsure whether your specific generations are currently being watermarked, you can check using the ElevenLabs Audio Detector. No ElevenLabs audio created prior to June 2026 carries a SynthID watermark.
Can the audio detector check audio from other AI platforms?
No. The ElevenLabs Audio Detector can only detect watermarks embedded by ElevenLabs. It will not detect SynthID watermarks applied by other platforms.
Do I have to be logged in to use the audio detector?
Yes, the Audio Detector requires signing in to your ElevenLabs account. The older AI Speech Classifier is available without login.