kyutai

The latter is developed by a single guy while Kyutai is funded with 150M€.

Tricky thing is personaplex forked some of the kyutai code and has not integrated the LoRA stuff they added.

My TTS/STT models are local by Kyutai and the voice agent's LLM between STT and TTS is used to determine some basic context: e.g.

I haven't tried them myself but the Kyutai has a couple projects that could fit.

Hello Gabriel from Kyutai here, maybe it's related to the way we chunk the text?

FYI my day job is at a frontier voice ai lab focused on hyper-realistic low latency TTS (we're a commercial spinoff of Kyutai called Gradium).

Gabriel from Kyutai here, we do support outputting wav to stdout.

I no specialist of the field at all, but in the context of Kyutai they explained their workflow a bit to make their speech to speech model.

I had a discussion about a year ago with a researcher at Kyutai and they told me their lab was spending an order of magnitude more compute in artificial data generation than what they spent in training proper.

The platform supports AI-powered meeting transcription using speaker diarization technology from the French startup pyannote, with real-time subtitle generation from AI research lab Kyutai expected to be added by summer 2026"[0].

There is no cross-attention in either Kyutai's STT nor in Voxtral 2, unlike Whisper's encoder-decoder design!

I have tried to take the fine tuning LoRA stuff from Kyutai/moshi-finetune and apply it to the personaplex code.

Curious how it compares to last week’s release of Kyutai’s Pocket-TTS [1] which is just 100M params, and excellent in both speed and quality (English only).