QwenAudio/Fun-Audio-Chat

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1009 stars · 105 forks observed · 2026-09-26

What it is AI-extracted, prompt v1, taxonomy v1, 2026-09-07, confidence 0.9

Fun-Audio-Chat is a Large Audio Language Model (8B) for natural, low-latency voice interactions, using dual-resolution speech representations and Core-Cocktail training. It ships model weights plus training and inference code supporting spoken QA, audio understanding, speech function calling, and instruction following.

Use cases

• build a low-latency voice chat assistant
• run spoken question answering on audio input
• add speech function calling to a voice agent
• understand and transcribe audio with an LLM
• fine-tune an audio language model
• serve an audio LLM with vLLM for faster inference

When to choose

• you need an open 8B audio-language model with strong speech benchmarks
• you want low-latency voice interaction with efficient 5Hz speech representations
• you need speech function calling or voice empathy in a voice assistant

When to avoid

• you only need plain text-to-speech or ASR without LLM reasoning
• you lack GPU resources for 8B model inference
• you need a small embedded or on-device audio model

Facets

library · maturity active

speech-recognition llm-inference llm-training tts machine-learning speech-processing large-language-models artificial-intelligence python audio-language-model voice-chat dual-resolution-speech function-calling vllm natural-language-processing gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="QwenAudio/Fun-Audio-Chat")