I presume you aready have your python environemnt set up, then you'd only need to install and import the modules.Thanks for posting that links. I don't use that kind of script language a lot. Do I just copy the section up top that begins with:
"from huggingsound import SpeechRecognitionModel"
and input the full input/output paths to the audio in the command prompt and it will do its work?
I don't see any info on having to download any files.
Also, it says it needs input of 16 kHz sample rate. So you would need to convert every audio file in Audacity to 16 kHz.
Is there a runtime limit to the audio? Vrew has a 2 hr limit and I know pyTranscriber doesn't like long files either.
Once again, thank you for letting us know there are more options out there.
I do remember someone else mentioning a 'new' method of extracting text from audio in THIS thread. Never found a more detailed explanation.
Both the input and output need pre and post work. And you're right, using Audacity for 16khz would do the job. I use ffmpeg for the same. I cut the input to 30min chunks. I'm still looking for a pre-trained (Japanese) tokenizer for preprocessing.
For pyTranscriber I usually split the input to 5min chunks before feeding it to pyTransciber.
PS. I also noticed that the results are better when audio is converted from streo to mono.
PPS. Somewhere I read that Vrew gives better results if the audio is converted to female voice, and the volume is increased 5x. I haven't tried it though.






