Qwen Audio
Speech recognition & audio understanding
Qwen-Audio-3.1-ASR
Multilingual speech recognition with non-streaming speaker diarization in Flash and Next, plus general audio understanding in Next.
30 languages16 Chinese dialects
Alibaba · Speech & audio research
Qwen Audio.
From listening and understanding to speaking and creating. Explore the Qwen-Audio series and our wider family of speech and audio projects.
Explore Qwen-Audio
Models · Audio demos · Technical reports
Real-time voice interaction
Qwen-Audio-3.1-Realtime
Explore spoken reasoning, tool use and natural interaction, alongside a persistent voice agent built as a separate runtime extension.
Spoken reasoningTool use
More speech & audio projects
Understand · Generate · InteractSpeech recognition · 3.0
Qwen-Audio-3.0-ASR
Multilingual and dialect recognition with streaming transcription, contextual recognition and hotwords.
Speech synthesis · 3.0
Qwen-Audio-3.0-TTS
Controllable speech synthesis, cross-lingual voice cloning and long-form generation.
Speech recognition
Fun-ASR
Speech recognition for multilingual, dialect and domain-specific audio.
Speech synthesis
CosyVoice 3
Explore the next generation of multilingual speech generation and voice cloning.
Voice conversation
Fun-Audio-Chat
Connect audio understanding and natural spoken interaction.
Speech synthesis
CosyVoice 2
Streaming synthesis, zero-shot voice cloning and multilingual generation.
Spoken interaction
MinMo
Multimodal language models for real-time, natural voice conversation.
Music generation
InspireMusic
Explore music generation, continuation and music creation examples.
FunAudioLLM · Understanding & generation
SenseVoice & CosyVoice
The original FunAudioLLM project: multilingual speech understanding, emotion and event recognition, voice generation and voice interaction demos.
Bring speech research into your application.
Explore inference, serving and application integration with FunASR. See each project for model availability and licensing.