HOWTO quality video/audio to text translation

Communities Science, Engineering, & Security
1

In the following I’m going to detail, how to take audio either from files or pulled from online videos and finally output the speech as text files.

Currently I’ve been saving a database with all the video essays and websites I see,
with text solution being part of that system.

To try this online with a video:

Else run it locally via python

import os
import pytube as pt
import whisper

# download mp3 from youtube video
yt = pt.YouTube("VIDEO_ADDRESS_HERE")
stream = yt.streams.filter(only_audio=True).first()
print("Getting text for video: {} - {}".format(yt.author, yt.title))

tempFile = "audio_english.mp3"

stream.download(filename=tempFile)
model = whisper.load_model("base")
result = model.transcribe(tempFile)

# write temp text file
print("writing output")
outFile = "output.txt"
with open(outFile, 'w') as f:
    f.write(result["text"])

# rename as final file
titleFile = "{} - {}.txt".format(yt.author, yt.title)
os.rename(outFile, titleFile)
2 Likes
2

Interesting. Is there something like this for non-YouTube vids?

3

There’s no dependency here,
so just swap out the pytube library for something else or directly give it an audio file.

4

This is pretty sweet. Led me to Whisper. I just wish there was an easier way to churn out into a subtitle format.

5

Have you tired this with the subtitle flags?

6

jesus, I was using whisper to transcribe and generate a .srt and used MKVToolNix to throw it all together. How in the world did I not know that was there.

At least I learned how to use whisper

1 Like