Video file to subtitles file

[MRAB]

On 2020-08-30 18:10, Christian Gollwitzer wrote:
> Am 30.08.20 um 17:25 schrieb MRAB:
>> On 2020-08-30 07:23, Muskan Sanghai wrote:
>>> On Sunday, August 30, 2020 at 11:46:15 AM UTC+5:30, Chris Angelico wrote:
>>>> I recommend looking into CMU Sphinx then. I've used that from Python.
>>>> The results are highly entertaining.
>>>> ChrisA
>>> Okay I will try it, thank you.
>>>
>> Speech recognition works best when there's a single voice, speaking 
>> clearly, with little or no background noise. Movies tend not to be like 
>> that.
>> 
>> Which is why the results are "highly entertaining"...
> 
> 
> Well, with enough effort it is possible to build a system that is more
> useful than "entertaining". Google did that, English youtube videos can
> be annotated with subtitles from speech recognition. For example, try
> this video:
> https://www.youtube.com/watch?v=lYVLpC_8SQE
> 
> Go to the settings thing (the little gear icon in the nav bar) and
> switch on subtitles, English autogenerated. You'll see a word-by-word
> transcription of the text, and most of it is accurate.
> 
There's not much background noise there; it takes place in a quiet room.

> There are strong arguments that anything one can build with open source
> tools will be inferior. 1) They'll probably have a bunch of highly
> qualified KI experts working on this thing 2) They have an enormous
> corpus of training data. Many videos already have user-provided
> subtitles. They can feed all of this into the training.
> 
> I'm waiting to be disproven on this point ;)
>