"One Click Transcribe" tool

[Purfview]

     8""""                                    8   8  8                                       8   8 8   8 8
     8     eeeee eeeee eeeee eeee eeeee       8   8  8 e   e e  eeeee eeeee eeee eeeee        8 8   8 8  8
     8eeee 8   8 8   "   8   8    8   8       8e  8  8 8   8 8  8   " 8   8 8    8   8        eee   eee  8e
     88    8eee8 8eeee   8e  8eee 8eee8e eeee 88  8  8 8eee8 8e 8eeee 8eee8 8eee 8eee8e eeee 88  8 88  8 88
     88    88  8    88   88  88   88   8      88  8  8 88  8 88    88 88    88   88   8      88  8 88  8 88
     88    88  8 8ee88   88  88ee 88   8      88ee8ee8 88  8 88 8ee88 88    88ee 88   8      88  8 88  8 88eee


Nothing to do. Read usage instruction at this link:
https://github.com/Purfview/whisper-standalone-win/discussions/337

Press any key to continue . . .

Some people don't understand power of CLI, they want useless GUI. So I added One Click Transcribe.bat to a pack.

How to utilize it:

1. Press the right mouse button on the "One Click Transcribe.bat" file.
2. Then in the context menu select Send To > Desktop (create shortcut) to create a shortcut on your Desktop.

Now you can drag and drop file or files or folder onto this shortcut to do automatic one click transcriptions.

"One Click Transcribe.bat" is just a text file, you can edit it with Notepad and add there the settings you prefer.

Note:

When task is to transcribe then language should be set to activate custom prompt presets, so if you leave it on auto then those presets won't be activated and transcription result may have worse punctuation ect.

Linux steps

1. Open terminal in the same folder where executable is.
2. Copy/paste the command below and press Enter, a shortcut should appear on a Desktop:

chmod +x faster-whisper-xxl

cat <<EOF > ~/Desktop/One_Click_Transcribe.desktop
[Desktop Entry]
Version=1.0
Type=Application
Name=One Click Transcribe
Comment=Process audio files with One_Click_Transcribe.sh
Exec=$(pwd)/One_Click_Transcribe.sh %F
Icon=audio-x-generic
Terminal=true
Path=$(pwd)
MimeType=audio/x-wav;audio/mpeg;audio/x-flac;audio/x-m4a;
Categories=Audio;Utility;
EOF

chmod +x ~/Desktop/One_Click_Transcribe.desktop

[vimbac]

thx, where is the file?

[Purfview]

In the r194.2 release.

[defaultsecurity]

Awesome. You have read my mind. I was just about to write such a script. Impeccable timing and the script is excellent. Thank you very much.

[dragokas]

How can I send pull request for this script? It has parsing bug.

[Purfview]

Post here about it and the fix.

[dragokas]

You're trying to make each argument to be safe by enquoting, but this construction is unsafe itself: for %%F in (%*)

Fixing:

• Crash on filenames having "(", ")" characters.
• Unable to process filenames with "!" characters.
• Possible current path issues.

Code:

@echo off
color 0A
echo.
echo.
echo      8""""                                    8   8  8                                       8   8 8   8 8     
echo      8     eeeee eeeee eeeee eeee eeeee       8   8  8 e   e e  eeeee eeeee eeee eeeee        8 8   8 8  8     
echo      8eeee 8   8 8   "   8   8    8   8       8e  8  8 8   8 8  8   " 8   8 8    8   8        eee   eee  8e    
echo      88    8eee8 8eeee   8e  8eee 8eee8e eeee 88  8  8 8eee8 8e 8eeee 8eee8 8eee 8eee8e eeee 88  8 88  8 88    
echo      88    88  8    88   88  88   88   8      88  8  8 88  8 88    88 88    88   88   8      88  8 88  8 88    
echo      88    88  8 8ee88   88  88ee 88   8      88ee8ee8 88  8 88 8ee88 88    88ee 88   8      88  8 88  8 88eee 
echo.
echo.                                                                                                            

:: Localize variables, enable extended syntax
setlocal enableextensions

:: Check if files were dropped onto the batch file
if "%~1"=="" (
    echo Nothing to do. Read usage instruction at this link:
    echo https://github.com/Purfview/whisper-standalone-win/discussions/337
    echo.
    pause
    color
    exit /b
)

:: Tool's path
set "dp=%~dp0"

:: Initialize file list
set "file_list="

:: Enquote all file paths
:while
for %%F in ("%~1") do set "file_list=%file_list% "%%~F""
shift
if "%~1" neq "" goto while

:: Start processing
"%dp%faster-whisper-xxl.exe" %file_list% -pp -o source --batch_recursive --check_files --standard -f json srt -m medium


pause
color
exit /b

[Purfview]

Thx. I'll add fixes in the next version.

[Soeroah]

Hi,

Thanks for this option. I tried to get it working with just my old batch file but it didn't want to play nice.

Could someone please help me tweak the one click batch file to output similar to what this older one did, plus speaker diarization?

C:\Users\David\Downloads\Faster-Whisper-XXL_r192.3.4_windows\Faster-Whisper-XXL\faster-whisper-xxl.exe "C:\Users\David\Documents\SoundRecordMp3\SoundRecordMp3\bin\Debug\Output" --language=en --model=medium --batch_recursive --beam_size=1 --output_format text --output_dir "C:\Users\David\Documents\SoundRecordMp3\SoundRecordMp3\bin\Debug\Output"

If I'm dragging the folder onto the batch file, I imagine I don't need the first line (where the .exe file resides) or the folder where the files I want are. The options, though, I'm a little confused where to put them in the batch. My gut is to replace

faster-whisper-xxl.exe %file_list% -pp -o source --batch_recursive --check_files --standard -f json srt -m medium

at the bottom of the text file with everything from --language on, adding the --diarize filter, but I'm not sure if the other stuff has changed between 192.3.4 and 245.2.0.0.

The core things I want are to output as a .text file in that particular output folder, and to have diarization enabled for at least english. Thank you

[Soeroah]

So the whole command should be

faster-whisper-xxl.exe %file_list% -o source -br -bs 1 -f text -m medium -l en --diarize pyannote_v3.0 -pp --check_files --standard --sentence --beam_size=1 --batched

Is that right?

And ah, I didn't know about copying the old models, but I don't know where those are installed or where to put them either and I'm up to 70% progress now

[Purfview]

Is that right?

If you use --standard then --sentence is redundant.
You are using --beam_size=1 twice as -bs 1 is the same thing.

I didn't know about copying the old models, but I don't know where those are installed or where to put them either

Models are at a "exe" file, in "_models" folder.

[Soeroah]

Thank you

I ran it on a conversation audio file to test it but it looks like the diarization wasn't very accurate - the same voice was speaker 02 and 01 at different points, two different voices were both speaker 03 and by the end it was just using speaker 00 for several speakers in a row. Is there a filter option that makes it take a bit longer to be more accurate or is it just a limitation of technology?

[Purfview]

Technology is not super accurate, you can test other diarize options. Use --help.
Anyway, these questions doesn't belong in this thread.

[Soeroah]

Oh, sorry.