Yeah, it is finally here! Still under development, but at least it is. The usage and docs are below.
To use the program make sure that you have:
- koboldcpp and of course the model. I recommend Vicuna 7B or WizardLM 7B the most.
- Python 3.6 at least (I suppose)
- ffmpeg
When it comes to video files, when it comes to downloading file from your desired platform of choice, it is the best to use yt-dlp. For .srt subtitles, it easiest to just grab YouTube subtitles from the video of course. But of course not all will have them. In that case, use OpenAI's Whisper or this port to .cpp for CPU usage ;) Put the .srt transcription in the "transcriptions" folder of course.
Don't change folder structure even if some types of files do not need to be placed in a certain folder. It is for the sake of keeping it future proof.
The CLI usage of the program is really not that complicated at all. You should understand it in no time by looking at the help message. Just run python ./main.py
But here's the most basic usage of the most important function of the program: python ./main.py aisummsrt ./transcriptions/invideo.srt lilo 28 ./invideo.mp4 finaloutput.mp4
For more advanced usage, head to the configuration file (fav2ucc.config.ini) and mess around a little with it.
When it comes to merging .srt files together into one file from few others, make sure that the files follow this format:
filename-p1.srt
filename-p2.srt
...
filename-pX.srtNonetheless, please keep in mind that the merging isn't perfect so recheck the file just in case in the splitted start-end sections per file.
- I keep providing some logs and sample video outputs here. Just search for "FAV2UCC" and read some descriptions of these videos.
- there is still a long way to go until the project will satisfy my own needs :)
- speaking of satisfying needs, if possible, I would want to try to make it as friendly as possible, both for devs and end users.
- would be good to make a actual GUI in the future for easier use. For now just CLI.