Recent content by t221152

I have no idea what you're asking me. Do you mean like confidence color coding ?

We've had these V2 vs V3 discussion a couple of times :P. So here are the numbers from actual tests on Japanese (higher is worse) BUT from what I understand, largeV2 is more forgiving when the audio quality is a lil worse cause it's trained on more messy audio. V3 is better on clear audio...

Not a fan of those settings tb Can't say I'm a fan of those settings. The VAD settings aren't supposed to be touched (besides threshold and VAD method pyannote vs silero). They are already optimized for most content. Yours are so far of the default, I'm really wondering how you got to these...

First pack has about 500 new subtitles from last time, more than 70% of them is probably from the last 2 months. This pack is ordered by label. You can see all the new files in the output.txt below the post. I've fixed some mistakes with the naming of HODV, if you find more big mistakes let me...

I found using deepseek through subtitle edit really bad, it was really slow as well. I've never ran deepseek locally tho. I rather use the full big model. There's a couple of sites that have been linked before where people do their translations. There's also a good script that has been linked...

Did he/she finally allow JAV ? I got banned 2 times from that one when it was still new for explicit content. Kinda weird he/she was looking into what you were uploading. It was clearly a manual ban, not some automatic trigger.

This seems to work, it's running right now. It's tight for me the VRAM usage with only 6GB. My latest gfx is AMD so :(, I have to run it on an older machine with a 1660 super. Last time it was weird cause torch.cuda.is_available() returned a false for me when I installed your requirements even...

For me it seems to be using the CPU. I have high CPU load and I see it using only 400MB of VRAM when I start it (It's usually between 3 to 4,5GB). 3% took 10 minutes which is even slow to be CPU. Copyright (c) 2005-2025 NVIDIA Corporation Built on Fri_Feb_21_20:42:46_Pacific_Standard_Time_2025...

Glad you liked it, I'm also incorporating a 2-pass system where we first do a little bit of context extraction to help out the whole translation process. It helps a lot with keeping certain terms and names the same. Example : In the first 600 lines the term 女将 (Okami) could be translated as...

https://github.com/yt-dlp/yt-dlp You gotta work with command line tho, and if it's embedded video you'll first need something that can catch the HLS streams (browser extension), cause just posting URL won't work into the command line. Supports multi-threading downloads as well. Most streaming...

I've switched to your script since the token usage is so much lower. Was using gptsubtitler but the token usage is like 5-6x. I did a few translations wasn't fully happy with the result, so I copied a lot of the prompt stuff gptsubtitler uses. I liked the result a lot more, but it might be a...

https://www.akiba-online.com/threads/post-your-jav-subtitle-files-here-jav-subtitle-repository-jsp-not-a-sub-request-thread.1466451/page-364#post-4904110 This cleaner removes stuff like that, as most hallucinations are always the same (like subscribe to my channel). It's the best one I've seen...

I did some testing with the flash and pro 2.5 model, 2 or 3 months ago, which in the AI universe is a century ago. It was a lil worse imo. The flash version is really cheap and very fast, which imo is really an option if you don't mind the slightly worse translation. Apparently the gemini pro...

Not sure if it's worth it, I think it's more when there's heavy background noise, like music (or a shower). But I am interested if anyone actually has tried it. The re-aligning of timings seems useful tho.

mdx_kim2 could help with that, it's a better voice extraction method than what faster-whisper uses (which is HT demucs I think ?). I use whisperXXL