> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tinfoil.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# Audio models

> Audio models available on Tinfoil for speech-to-text transcription, audio understanding, and text-to-speech.

> [card]
>
> ![OpenAI](https://mintcdn.com/tinfoil/7kpqELCdP4WIVCil/images/model-icons/openai.png?fit=max\&auto=format\&n=7kpqELCdP4WIVCil\&q=85\&s=1b9cbe9cd3cbfe37d282867e0d448c69)
>
> Whisper Large V3 Turbo
>
> [`whisper-large-v3-turbo`](#whisper-large-v3-turbo)
>
> **Parameters:** 809M
>
> **Quantization:** None (served in FP16)
>
> **Capabilities:** Speech-to-text transcription of uploaded audio files
>
> **Audio Format:** File upload via the `/v1/audio/transcriptions` endpoint with `model=whisper-large-v3-turbo`
>
> **Languages:** More than 90 languages
>
> **Best for:** Transcription, captioning, voice interfaces, and multilingual audio processing
>
> **Model weights:** [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo)
>
> **Configuration repo:** [tinfoilsh/confidential-realtime-models](https://github.com/tinfoilsh/confidential-realtime-models)

> [card]
>
> ![Mistral](https://mintcdn.com/tinfoil/vfMAM73wcpT2SrpR/images/model-icons/mistral.png?fit=max\&auto=format\&n=vfMAM73wcpT2SrpR\&q=85\&s=c629064347c0c847d788dc2b2e0c1e10)
>
> Voxtral Small 24B
>
> [`voxtral-small-24b`](#voxtral-small-24b)
>
> **Version:** 2507 release
>
> **Parameters:** 24B
>
> **Quantization:** None (served in BF16)
>
> **Capabilities:** Speech-to-text transcription, audio Q\&A, summarization, translation, voice-triggered function calling
>
> **Audio Duration:** Up to 30 minutes (transcription) or 40 minutes (understanding)
>
> **Audio Format:** Supports .mp3 and .wav files
>
> **Languages:** English, Spanish, French, Portuguese, Hindi, German, Dutch, Italian
>
> **Best for:** Speech transcription with automatic language detection, answering questions from spoken input, generating summaries from audio, and triggering functions from voice commands
>
> **Model weights:** [mistralai/Voxtral-Small-24B-2507](https://huggingface.co/mistralai/Voxtral-Small-24B-2507)
>
> **Configuration repo:** [tinfoilsh/confidential-voxtral-small-24b](https://github.com/tinfoilsh/confidential-voxtral-small-24b)
>
> > [!NOTE]
> >
> > **Audio + Text:** Built on Mistral Small 3.1 foundation, combining speech processing with strong text capabilities including function calling from voice commands.

> [card]
>
> ![Mistral](https://mintcdn.com/tinfoil/vfMAM73wcpT2SrpR/images/model-icons/mistral.png?fit=max\&auto=format\&n=vfMAM73wcpT2SrpR\&q=85\&s=c629064347c0c847d788dc2b2e0c1e10)
>
> Voxtral TTS
>
> [`voxtral-tts`](#voxtral-tts)
>
> **Version:** 2603 release
>
> **Parameters:** 4B
>
> **Quantization:** None (served in BF16)
>
> **Capabilities:** Converts text into natural-sounding spoken audio
>
> **Audio Format:** Returns generated audio via the `/v1/audio/speech` endpoint with `model=voxtral-tts`
>
> **Best for:** High-quality voice output for assistants, narration, and accessibility, kept inside a secure enclave
>
> **Model weights:** [mistralai/Voxtral-4B-TTS-2603](https://huggingface.co/mistralai/Voxtral-4B-TTS-2603)
>
> **Configuration repo:** [tinfoilsh/confidential-realtime-models](https://github.com/tinfoilsh/confidential-realtime-models)

> [card]
>
> ![Qwen](https://mintcdn.com/tinfoil/7kpqELCdP4WIVCil/images/model-icons/qwen.png?fit=max\&auto=format\&n=7kpqELCdP4WIVCil\&q=85\&s=ceccb0e82dd42dae1af15f79d28fe557)
>
> Qwen3 TTS
>
> [`qwen3-tts`](#qwen3-tts)
>
> **Parameters:** 1.7B
>
> **Quantization:** None (served in BF16)
>
> **Capabilities:** Text-to-speech synthesis with custom voice support
>
> **Audio Format:** Returns generated audio via the `/v1/audio/speech` endpoint (the default model for that endpoint)
>
> **Best for:** Lightweight, low-latency speech generation
>
> **Model weights:** [Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice](https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice)
>
> **Configuration repo:** [tinfoilsh/confidential-realtime-models](https://github.com/tinfoilsh/confidential-realtime-models)

> [card]
>
> ![Mistral](https://mintcdn.com/tinfoil/vfMAM73wcpT2SrpR/images/model-icons/mistral.png?fit=max\&auto=format\&n=vfMAM73wcpT2SrpR\&q=85\&s=c629064347c0c847d788dc2b2e0c1e10)
>
> Voxtral Mini Realtime
>
> [`voxtral-mini-4b-realtime`](#voxtral-mini-4b-realtime)
>
> **Version:** 2602 release
>
> **Parameters:** 4B
>
> **Quantization:** None (served in BF16)
>
> **Capabilities:** Streaming speech-to-text transcription, with partial transcripts appearing word-by-word as audio arrives (\~480ms behind the speaker)
>
> **Audio Format:** PCM16 mono over WebSocket (16kHz native; 24kHz accepted on the OpenAI-compatible endpoint and resampled in-enclave)
>
> **Languages:** English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian, Dutch
>
> **Best for:** Live dictation, voice interfaces, and any UI that should show words as they're spoken rather than after the recording ends
>
> **Model weights:** [mistralai/Voxtral-Mini-4B-Realtime-2602](https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602)
>
> **Configuration repo:** [tinfoilsh/confidential-realtime-models](https://github.com/tinfoilsh/confidential-realtime-models)
>
> > [!WARNING]
> >
> > This model is picky. Realtime transcription streams over a WebSocket, which currently can only be verifiably connected to using the [`tinfoil-js`](https://github.com/tinfoilsh/tinfoil-js) SDK, and only in Node, so not in browsers. If this is a problem, [contact us](mailto:contact@tinfoil.sh).
>
> > [!NOTE]
> >
> > **Streaming by architecture:** Transcripts stream incrementally, are append-only, and are never revised. Served over the WebSocket `/v1/realtime` endpoint, including an OpenAI Realtime API-compatible mode. See the realtime section of the [processing audio guide](https://docs.tinfoil.sh/guides/processing-audio.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.