Mini-Omni

Speech in, speech out, skipping the ASR-LLM-TTS relay race.

When to use it

Use this when you're prototyping real-time voice conversation and want one streaming model instead of three chained ones.

Text-to-speech, speech-to-text and audio models under one MLX package.

A batched, quantized Whisper reimplementation chasing raw transcription speed.

Prompt-controllable speech: describe the voice you want in plain English.