Skip to main content
Quail Multi Speaker processes audio for machine listening, including speech-to-text (STT) and voice agents. The multi-speaker models target noise and reverberation while retaining speech from multiple participants. Use Quail Multi Speaker when every speaker matters, such as a meeting transcription or speakerphone application. If your application should follow one primary speaker and suppress competing speech, evaluate Quail Voice Focus.

​
Choose a variant

The small and large variants give you alternatives to compare for quality and resource use. Measure both with your recordings, STT provider and target hardware. The model reference lists release-specific rates and block sizes.

​
Evaluate the result

Compare the original and enhanced audio using the same STT model, transcription settings and labeled recordings. Track word error rate (WER), insertions, deletions and task completion. Enhancement does not guarantee an improvement for every recording or STT provider.
Quail Multi Speaker is optimized for machine listening. Some noise or reverberation can remain, and the output may differ from what a human listener considers natural. Use Rook Multi Speaker when human listening is the primary goal.
Follow the STT tuning guide and SDK quickstart to evaluate a complete integration.