2

it looks like all the questions about this topic are at least a few years old, and technology advances quickly, so I'll go ahead and ask. At this time (Feb 2024), what are the top and/or the best voice and speech recognition software options?

I am thinking of something that will recognize words and sentences. I am in the dreaming/thinking stage of a project. In one part of the project, the software would ask the user to repeat a couple of words, a sentence or two, or even maybe a paragraph or two. I would want the software to be able to be fairly certain (within an acceptable margin of error) as to whether the respondent said the phrases they were supposed to say.

I am interested in both free and paid options, subscription or one-time payment, open-source or closed source, AI-backed or non-AI backed. I am looking for as many options as possible. Again, I am just in the thinking stage, so there is no imagined budget. I am just trying to get info to see if my project is feasible.

This is only one part of the project. I am reluctant to share the details of the project itself, in case I want to pursue it. If I decide not to pursue it, I might share the details at that point for anyone who might want to pursue it.

Thanks for all your suggestions!

Improve this question
2

2 Answers 2

Reset to default
1

Whisper is very popular AI-based one right now, with various implementations, such as open source Whisper (Python) or Whisper.cpp.

It can use bunch of different models, which can be found e.g. here for the C++ version. Of course the larger the models, the tougher/longer but more accurate the computations will be. Note that there are also various versions of the software, e.g. the default C++ one relies on processor, but you can potentially also use one that utilizes CUDA.

Improve this answer
3
  • Do you know if there exist a wrapper around these libraries? I am aware of the command line client but it is very limited compared to what the libraries offers.
    – d-b
    Commented Oct 23 at 22:40
  • Limited in what way? YYou should ask for what specifically you're looking for
    – Destroy666
    Commented Oct 23 at 22:49
  • E.g., you can't set the temperature but have to use workarounds in the CLI-client. I want to use multiple preferences, depending on the circumstances.
    – d-b
    Commented Oct 23 at 23:53
0

I am using Vosk, which works reasonably well for my native language (German) and not only English (probably better than German). I use it for transcribing Youtube videos.

For German language, it sometimes has problems when we mix English words into the German text. That's common in software development and a bit annoying. YMMV, depending on the language you want to use.

Vosk is gratis and Open Source with the permissive Apache 2 license (see Github).

You didn't specify any hard requirements, so I can't judge on those.

Improve this answer

Your Answer

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

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