feat(mistral): add voxtral TTS support#5245
Conversation
|
It's worth mentioning that both |
|
Hello @jeanprbt I just realized I did the same PR as you did 2 hours ago, sorry for that. Just wondering where you found the labels for the TTS ? I just found this one on the doc: "voxtral-tts-2603" |
|
Hi @fabitokki,
I found the labels here. You can rely on this PR for official Mistral support of Voxtral TTS, no worries! |
|
Would really love to have that. This might be some real competition for 11labs. When can we expect to have that available on the pip repository? |
34e819b to
143fec3
Compare
As soon as this PR is merged I guess! Looking forward too ;) |
|
@fabitokki Done, I just edited the description. Thank you! |
|
@jeanprbt Cool, I closed my PR to avoid duplication. Have a good day! |
Summary
livekit-plugins-mistralaiplugin using Mistral's newvoxtral-mini-tts-2603modelDetails
TTS implementation (
tts.py)ChunkedStream-based synthesis viaclient.audio.speech.complete_async()with streamed outputaudio_datafrom the API responseAPITimeoutError,APIStatusError, andAPIConnectionErrorpcmaudio format is default as it's fastest, requires additionalfloat32->float16conversionModels & voices (
models.py)TTSModelsliteral:voxtral-mini-tts-2603,voxtral-mini-tts-latestTTSVoicesliteral with 28 voice presets (Jane, Marie, Oliver, Paul)This would solve issue #5247.