Trying to generate a subtitle file

[archz2]

I am using whisper in google collab to generate subtitle file of an mp3 file. I am getting this error. What should I do?

[EtienneAb3d]

Try this code:
https://github.com/EtienneAb3d/WhisperTimeSync/blob/main/transcribe.py

[archz2]

Thanks. And if the audio is in Korean and I want to generate english subtitle file?

[EtienneAb3d]

Try:

result = model.transcribe(mp3,language="en")

[archz2]

Okay actually I used a simpler code.

!whisper finaOctaneMaterials.mp3 --language=ko

result = model.transcribe(mp3,language="en") 

It is working now.

I have three questions.

1. How do I make it use GPU so that it runs faster?
2. It is directly generating subtitles in Korean instead of English even though I used 'result = model.transcribe(mp3,language="en") '. What should I do so that it directly generates them in English?
3. It is using the 'large' model for Korean also, right?

[archz2]

Okay what lines do I add to make it use the GPU?

P.s. Got it.

I used the following code now.

import sys
from pathlib import Path
import whisper
from whisper.utils import WriteSRT

modelSize = "large"

if __name__ == "__main__":
    mp3 = sys.argv[1] #"data/finaOctaneMaterials.mp3"
    
    if len(sys.argv) >= 2 :
        modelSize = sys.argv[2]
    
    print("Loading Whisper model: "+modelSize+" ...")
    model = whisper.load_model(modelSize)
    
    print("Transcribing: "+mp3+" ...")
    result = model.transcribe(mp3,language="en") 
    
    print(result["text"])
    
    print("Saving: "+mp3+".srt ...")
    p = Path(mp3)
    writer = WriteSRT(p.parent)
    writer(result, mp3)

I got this error.

[EtienneAb3d]

Your notebook config looks like mine.

It should display:

Loading Whisper model: large

If you are executing the code directly, you should perhaps remove the "main" line.

[archz2]

I don't understand. What should be the new lines of code now? I ran your code by copy pasting in Google Collab.

[EtienneAb3d]

With the "main" line, this code is written to be used as a command line, like:

python3 transcribe.py data/KatyPerry-Firework.mp3 large

To use it directly in interactive mode, you should remove the "main" line, like:

import sys
from pathlib import Path
import whisper
from whisper.utils import WriteSRT

modelSize = "large"

mp3 = "data/KatyPerry-Firework.mp3"

print("Loading Whisper model: "+modelSize+" ...")
model = whisper.load_model(modelSize)

print("Transcribing: "+mp3+" ...")
result = model.transcribe(mp3)

print(result["text"])

print("Saving: "+mp3+".srt ...")
p = Path(mp3)
writer = WriteSRT(p.parent)
writer(result, mp3)

[archz2]

Okay. I used this code

import sys
from pathlib import Path
import whisper
from whisper.utils import WriteSRT

modelSize = "large"

mp3 = "data/Scene1.mp3"

print("Loading Whisper model: "+modelSize+" ...")
model = whisper.load_model(modelSize)

print("Transcribing: "+mp3+" ...")
result = model.transcribe(mp3)

print(result["text"])

print("Saving: "+mp3+".srt ...")
p = Path(mp3)
writer = WriteSRT(p.parent)
writer(result, mp3)

and got the following error.

[EtienneAb3d]

"No such file or directory"
Please, try to investigate by yourself before asking for our time for such basic coding problem: the path you provided is not the good one. As I see on your left panel, your file is not in the sub-directory "data/".

[archz2]

When the Korean subtitles got generated my file was in 'sample_data' folder. I did not create any data folder. How do I ensure that the file inside 'sample_data' folder is used?

I tried an alternative. My file titled '02_02_Workflow.mp3' is in a folder called WhisperAudio in my Google drive. If I use the following address

"content/drive/MyDrive/WhisperAudio/02_02_Workflow.mp3"

I get the same error.