Skip to content

Feature Proposal: Add FunASR ASR Integration for Voice-Enabled Qwen2 Applications #1864

Description

@LauraGPT

Qwen2 supports multimodal input, and integrating speech-to-text would enable voice-enabled Qwen2 applications. FunASR (17.8K+ stars, https://github.com/modelscope/FunASR) is from the same Alibaba/ModelScope ecosystem and provides:

  • SenseVoice: Ultra-fast multilingual ASR (50x faster than Whisper-large, strong CJK + Cantonese)
  • Paraformer: Non-streaming ASR with timestamps and punctuation
  • Fun-ASR-Nano: Lightweight streaming ASR for edge deployment
  • OpenAI-compatible API: POST /v1/audio/transcriptions

Since FunASR and Qwen2 are both ModelScope projects, they naturally complement each other — FunASR handles speech input while Qwen2 handles text understanding and generation. Together they form a complete voice-to-text-to-response pipeline, all self-hosted.

Would it be useful to add FunASR integration examples in Qwen2 docs?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions