On-device voice AI — MLX (Apple) and CUDA (NVIDIA)

Streaming speech recognition, neural voice synthesis with visemes, and tool-calling chat — running locally on the active backend (see the pill in the header).

Dudoxx Omni Intelligence · MLX + CUDA · FastAPI · Next.js 16

Service status

STT…TTS…LLM…

Speech-to-Text

Whisper-class streaming ASR over WebSocket. Word-level timestamps, multilingual.

Open playground

Text-to-Speech

Qwen3-TTS and Kokoro voices with Preston-Blair visemes for lipsync.

Open playground

Local Chat

Qwen3-4B with OpenAI-compatible tool calling. ~104 tok/s on M-series.

Open playground