Skip to content

Feature request: Support FunASR/SenseVoice as local ASR backend #567

@LauraGPT

Description

@LauraGPT

Feature Request

I'd like to suggest adding SenseVoice as a local ASR backend option for OpenLess.

Why this benefits OpenLess users

OpenLess's "hold key, speak, release" workflow demands fast, accurate local transcription with minimal latency. SenseVoice-Small is an excellent fit:

  • 234M parameters — lightweight enough for real-time local inference on consumer hardware
  • Fully offline — no API calls, no network latency, perfect for the instant-response feel OpenLess provides
  • 50+ language support — works globally without switching models
  • Emotion/event detection — can detect laughter, applause, and speaker emotion, potentially useful for AI text polishing context
  • High accuracy — comparable to Whisper-large but at a fraction of the compute cost

Technical advantages

  • Processes audio at 170x realtime speed — near-instant transcription on release
  • Single model handles multilingual transcription without language detection step
  • Available via simple Python API (pip install funasr)
  • No GPU required for inference on short utterances (typical for voice input)
  • Apache 2.0 licensed, no usage restrictions

Links

This could give users a fast, private, and accurate alternative for local speech-to-text.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    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