Fun-Audio-Chat
Alibaba Cloud's ultra-low-latency voice interaction model — empathetic, full-duplex
Alibaba Cloud's ultra-low-latency voice interaction large model; empathetic, with full-duplex conversation.
Voice AI: lifelike TTS, accurate ASR, and real-time voice agents.
Key features
- Large Audio Language Model for natural low-latency voice interaction
- Dual-resolution speech representations cut compute while keeping quality
- Spoken QA, audio understanding, speech function calling, voice empathy
- Core-Cocktail training preserves strong text LLM capabilities
Good for
- Developers building voice chat assistants
- Teams benchmarking speech-to-text and speech-to-speech tasks
Stack & tags
PythonPyTorchvLLMNode.jsvoice-chataudio-llmspeech-to-speechspoken-qalow-latencyopen-weights
Quick start
export PYTHONPATH=$(pwd)
python examples/infer_s2t.pyFrom the project README on GitHub.