Re: Translating captions

[Jean-Christophe Helary]

> On Dec 9, 2022, at 11:58, Esteban Ordóñez <quiliro@riseup.net> wrote:
> 
> El 2022-12-08 17:12, Jean-Christophe Helary escribió:
> 
>>> If it takes so long to transcribe, wouldn't it be easier to transcribe
>>> manually?
>> 
>> Manual transcription requires around 10 minutes per minute of talk.
> 
> I calculate that it takes (someone who knows how to touch-type) about 2
> seconds per word to transcribe, including corrections during and after. 
> So, if the speaker speaks at 1 second per word, transcription would take
> 2 minutes per minute of talk.

Ten minutes is the accepted value for commercial transcription services. It 
also depends on the density of the talk, the experience of the transcriber, etc.

Average speaking rates depend on the material, of course:
https://virtualspeech.com/blog/average-speaking-rate-words-per-minute

150 wpm would be 2.5 wps.

>> Whisper is way faster than that.
> 
> Probably it depends on the processor.  GPU processing is the one in
> charge of these repetive tasks.  I am not sure how efficient it is to
> use free software for GPUs.

On my 7-year-old Macbook Pro, it’s fast enough that I can spend my time working 
on things the computer can’t do in my place :-)

-- 
Jean-Christophe Helary @jchelary@emacs.ch
https://traductaire-libre.org
https://mac4translators.blogspot.com
https://sr.ht/~brandelune/omegat-as-a-book/