Compatibility with MistralAI's Voxtral for Audio Transcription

[flefevre]

Hi Litellm team,

Thanks for your great work on this project!

I wanted to ask if the litellm.audio.transcribe() method is compatible with MistralAI's new Voxtral model for audio transcription, recently announced here: https://docs.mistral.ai/capabilities/audio/

From the documentation (https://docs.litellm.ai/docs/audio_transcription), I see that audio transcription is supported, but it's not clear whether Voxtral is currently integrated or can be used via the standard interface.

Is Voxtral currently supported?

If not, are there plans to support it soon?

Would it be possible to manually configure a provider to call Mistral’s API directly using Litellm?

Any clarification would be appreciated. Thanks again!

Best,
François

[wm-ek]

+1 any updates on this?

tried to integrate without success yet

[ishaan-jaff]

Hi team, great request - i've added it to out backlog

[wm-ek]

@ishaan-jaff thanks.

if someone needs a workaround:
currently i am using it as passthrough. not as nice for tracking, but bether than notihg

and set the Authorization: Bearer KEY header

[ishaan-jaff]

that looks like OCR tho ? The ask is for voxtral

Mistral OCR live here: https://docs.litellm.ai/docs/ocr

[wm-ek]

sorry, I mixed things up, you're right. with ocr i had problems getting it working on openwebui, so i did it that way.