I have hundreds of hours of audio files in English that I need to transcript to the same language. Do you know a speech-to-text software that I can use to do it automatically?

  • Even when its automatic transcription will not be perfect, it will be good enough for my use case
  • Support English language at least
  • Open source
  • It should runs on Linux, but other platforms are okay
  • The audio files are on MP4, but I can convert them to a different format if it's necessary
  • The result should be simple text files
  • Privacy is important, so I can't use an online service

I imagine it like an application which I could use it like this:

$ transcriber --input 001.mp4 --output 001.txt

With something like this, it will be easy to write a script to transcribe all the audio files.


  • 1
    Is this a one-off transcription run? You might be better off with an online service, e.g. IBM's Watson Speech-to-Text: not open source, obviously, but inexpensive at 2 cents per minute and the first 1000 minutes free.
    – Rup
    Commented Jul 19, 2016 at 20:22
  • I didn't know that. It's very interesting, but the idea about have it Open Source is to save some money and, the most important, I will not have to worry about privacy issues, because it will run on my computer.
    – Mario S
    Commented Jul 20, 2016 at 22:55
  • 1
    A Google search for speech to text software brings me e.g. to Top Speech Recognition Software Products, but I don't know how independent/good this is.
    – user416
    Commented Aug 4, 2016 at 10:18
  • Possible duplicate of Interview transcription software
    – Tom
    Commented Aug 19, 2016 at 19:27
  • 2
    github.com/openai/whisper works great for this Commented Oct 14, 2022 at 18:16

3 Answers 3


This is going to sound somewhat weird, but you can upload it to Youtube (assuming its not copyrighted material) and have it add its auto CC, and download that file.

Far from perfect, but a possible solution.

  • 1
    Privacy is important. I updated the question. Thanks anyway,
    – JonDoe297
    Commented Jul 25, 2016 at 21:22
  • I understand. Youtube can still be considered as an option, the file can be privated (granted, you'd be uploading it online). I'm actually curious if you do find a better solution Commented Jul 26, 2016 at 18:20

OpenAI's Whisper is pretty slow, but has state-of-the-art performance. There currently is no system that has higher accuracy for general purpose transcription.


I recommend using https://transcribe.wreally.com. Neither your audio file, nor your transcription is transmitted to their website, so your privacy requirement is satisfied. You can use it offline but you have to use it in the Chrome browser. (Unfortunately it does not satisfy your open source requirement.)

It is similar to youtube's transcription process, but better.

  • How is this open source? Can you link to an open source license they provide? Commented Aug 21, 2016 at 17:54
  • 1
    @ThomasWeller - I guess you missed where I wrote "Unfortunately it does not satisfy your open source requirement." JonDoe's request involves many constraints. I was clear about one of them not being satisfied by my suggestion. Commented Aug 22, 2016 at 0:11
  • Alright. I think it's ok since you explicitly mention it. If that sentence was in your original post, I must have missed it. Sorry. Commented Aug 22, 2016 at 20:23
  • @aparente001 What I was looking for is an automatic transcription software. Not only transcription. It was my fault the question was not as clear as I thought. Anyway, I updated it and added a new comment. Thanks.
    – JonDoe297
    Commented Aug 23, 2016 at 0:39
  • This is a very difficult problem. It might be doable if it's one speaker, speaking very clearly, with excellent acoustic conditions. The Wreally source I cited does offer this but I haven't tested it. Although it's not open source, this might be a way for you to do a reality check, i.e. check feasibilty.of your idea. Commented Aug 27, 2016 at 15:44

Not the answer you're looking for? Browse other questions tagged or ask your own question.