Is there a means to operate this by means of the command line as simply just as common Whisper? I have a bunch of problems with looking to do matters by means of standard Python code. I face far more VRAM challenges, and many others and it won't spit out all of the various subtitle formats quickly.
Has any person accomplished that even though? When you have a high-quality tuned product, Whisper is vastly harder to operate so far as I'm able to convey to.
I just love this oldie Mom-Son themed Incest JAV. I had downloaded a Sub for it quite a few years in the past but didn't go any even further because of the movie top quality but I recently utilized a "AI Enhancement" tool to the movie and made a decision to move forward.
The downloaded Sub was, I believe, made from challenging-coded Subbed online video clips by Spikespen. Due to the fact Spikespen only Subbed clips there were numerous gaps in dialog when placed on the whole Film. I made use of WhisperJAV0.7 to try to fill these gaps and I also tried to scrub it up a tiny bit and re-interpreted many of the meaningless/ "lewd-much less" dialog.
JUR-037 Eng Sub. Even though my bland partner was on a company trip, I was fucked and impregnated by his horny stepson.
Thanks a great deal for sharing these with us. I know I speak For lots of who respect all you and the opposite subbers add On this Discussion board.,
Once more, I don't comprehend Japanese so my re-interpretations may not be completely correct but I seek to match what is occurring from the scene. In any case, appreciate and allow me to know very well what you believe.
compression_ratio_threshold: two.4 default. Some measure of the amount of the transcription really should be distinct and not only the same line repeatedly in a method that may compress too nicely. I believe? No idea the way to intuit this.
Appears like looking to get that working fully broke my capability to operate common Whisper, so I guess I am from the game for your foreseeable future. Appears like library pathing is really a disaster for whisper-ctranslate2, so you need to soar via a lot of hoops.
2. Nearly anything that wasn't observed by means of R18.dev database, there were about 1600 subtitles remaining from the primary script, was then initial scraped vs Javguru and Next Javmost.
That is a few of the worth in running Whisper numerous occasions for every here job. One of these will pick up an real translation rather than (Talking in international language).
It might not totally be a true translation since I don't know Japanese or Chinese. (Guess I really should always incorporate this disclaimer!). In any case enjoy and allow me to know very well what you're thinking that.
Therefore, the vast majority of my advice In this particular thread will likely be for maximizing the standard of the Medium models.
A: I can't help you, sorry. Not one of the ways are really hard but there might be compatability concerns that need possibly seriously being aware of what you are executing or simply restarting your complete installation system. Just be sure you be sure you are employing Cuda if you wish to leverage your GPU.