ElevenLabs releases Cue, a forced alignment model for character-level audio timing
Article timeline
The articles in this story. Times are ET.
ElevenLabsOfficialElevenLabs releases Cue, a forced alignment model for any audio lengthAIElevenLabs says its new forced alignment model, Cue, places every character on the audio timeline more precisely than any aligner it tested. On real-world audio, Cue was more accurate than every open-source aligner tested across 29 languages, and it aligns an hour of audio in about half a second on one GPU. Cue powers ElevenLabs Speech to Text and is also available through the Forced Alignment API.
- ElevenLabs BlogOfficialElevenLabs' Cue aligns transcript characters to audio in a single pass
AIElevenLabs introduces Cue, a forced alignment model that places every character in an audio file on a timeline and powers its Scribe speech-to-text models. ElevenLabs says Cue beat every aligner it tested on real-world recordings in all 29 benchmark languages. The company reports average timing errors of 179 milliseconds on full recordings and aligns an hour of audio in about half a second on its production GPU.