Pure C implementation of Voxtral-4B-TTS-2603

ComfyUI nodes for Mistral's Voxtral-4B text-to-speech model with direct PyTorch inference. Supports 20 preset voices across 9 languages on CUDA, MPS, and CPU.

Research about Voxtral, its codebooks and an attempt to reconstruct codes for a known audio

No description available

Colab notebook for Voxtral TTS

Production-grade text-to-speech server. Pure C/C++ inference with voxtral-tts.c. OpenAI-compatible API. Sibling to openasr.