ElevenLabs released open-source AI Speech Classifier for detecting ElevenLabs-generated audio
Jan 30, 2024On January 30, 2024 - days after the New Hampshire fake-Biden robocall - ElevenLabs released an open AI Speech Classifier tool allowing journalists, election officials, fact-checkers and consumers to determine whether an audio clip was likely generated by ElevenLabs voices. The detector was made available without API gating or rate limits. Independent evaluation in 2024 found accuracy of approximately 95% on ElevenLabs-generated audio but limited applicability to audio from competing voice-clone systems.