AI Speech Classifier
The AI Speech Classifier analyzes audio characteristics to estimate ElevenLabs origin. It remains accessible as a legacy product and is also used within the newer Audio Detector.
Statistical detection and watermarking
The classifier provides a statistical estimate. The Audio Detector first checks for an ElevenLabs watermark, then invokes the classifier if no watermark is found. Watermark detection is a different method, not a new version of the classifier.ElevenLabs: Audio Detector and legacy classifier
Practical guidance. Treat a classifier result as evidence with limits. A negative result does not establish that audio is human speech or that it came from another specific provider.ElevenLabs: Audio Detector and legacy classifier
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Provider product |
| Scope | Older ElevenLabs speech; not a universal AI-audio detectorElevenLabs: deepfake detection methods and limits |
| v3 limitation | Cannot reliably classify Eleven v3 audioElevenLabs: deepfake detection methods and limits |
Related models
- Eleven Multilingual v1
ElevenLabs · Eleven speech
Retired first multilingual speech model, covering English and seven additional languages.
Launch date RetiredAPI · Provider product - Eleven Monolingual v1
ElevenLabs · Eleven speech
Retired original English speech model that preceded the multilingual family.
Launch date not confirmedRetiredAPI · Provider product