I'm looking for a speech recognition utility which generates subtitles. That is, given an audio file (or audio-and-video file), attempts to decipher the segments of spoken text in the audio, and generates a best-effort subtitle file with the text at different segments.

In this age of neural-network-learning, I'm betting this is no longer such an out-of-reach problem. Plus, it seems Google seems to have something like this on YouTube, albeit proprietary.

Required features:

  • gratis
  • Works on Linux
  • Supports at least one or two common audio (or A/V) formats
  • Must try to differentiate sentences / speech by different people - not just dump a long bunch of many paragraphs of text.

Desired features

  • libre
  • actively maintained
  • high accuracy rate
  • Supports many formats
  • Multi-platform
  • Apple's built-in Dictation utility can do the recognition part of this & transcribe to a text file, though it's a bit of a hack, as that's not what it was designed for. See jeffgeerling.com/blog/2022/… for a guide. It's been part of the OS about 10 years & can work offline as well as online.
    – Tetsujin
    Commented May 6, 2023 at 8:20
  • Modern browsers may have speech recognition API. See developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition if you are interested in programming a bit.
    – Marcel
    Commented Oct 22, 2023 at 19:48
  • @Marcel: 1. This is SR.SX, not SO... 2. The Web API specification is just that, a spec, not a product (nor a service). As far as I can tell, anyway.
    – einpoklum
    Commented Oct 22, 2023 at 20:41


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.