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

# Whisper

> Local speech-to-text on ESP32 voice AI toys with Whisper large-v3-turbo, running on Apple Silicon through MLX.

OpenToys transcribes your toy's microphone audio with **Whisper large-v3-turbo**, loaded through [`mlx-audio`](https://github.com/Blaizzy/mlx-audio).

| | |
| - | - |
| **Model** | `mlx-community/whisper-large-v3-turbo-asr-fp16` |
| **Runs on** | Apple Silicon via MLX |
| **Languages** | Multilingual (English, Chinese, Spanish, French, Japanese, Korean, Portuguese, German, Italian and more) |

## How it works

1. The ESP32-S3 captures audio from an I2S microphone (INMP441) and streams it to your Mac over a websocket.
2. Voice activity detection finds the end of your sentence.
3. Whisper transcribes the utterance and hands the text to the [LLM](https://elatoai.com/docs/local/llm/overview.md).

> [!NOTE]
>
> The model downloads on first setup into the app data directory. After that it works offline.


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