no need for comparison HEVC is better than Hi10 cope hi10 suckers

While there is considerably less banding and aliasing in Vardë's encode, you can see detail loss on the notebook in comparison 4 and the curtain in comparison 7. I'll just keep Commie's version.

well less banding vs. less detail in darker scenes is just the tradeoff you have to make unless you want to put in insane effort with scenefiltering/masking.

@Anon2954367 I sadly would agree. Overall the line art was washed out by a bad upscaler that caused blurrier line art (if that was used) or maybe the deband used was a bit too much (from experience) or something I dunno. Can really tell on screenshot comparison 1/9 on the green shirt and on the yellow brick stuff by the shoulder and so forth. Seeing on the website doesn't really present how much the line art is effected. View it on some photoshop program. Encoding settings https://animetosho.org/file/padpa-wotaku-ni-koi-wa-muzukashii-oad-02.880406 https://animetosho.org/file/commie-otaku-ni-koi-wa-muzukashii-oad-2.880276

commie on pic 2 in dark area just compressed mess lol

you can see varde's filtering [here](https://github.com/Ichunjo/encode-scripts/blob/master/Other%20groups/Natsumi%20no%20Sekai/Wotaku%20ni%20Koi%20wa%20Muzukashii%20OAD/wotakoi_ova_2.py#L79) meanwhile mine is just maa and neo_f3kdb(preset="low") lol

I took a "quick" look at this, Commie's, and UCCUSS-1Selxo's release and UCCUSS encode wins with a fair margin if you ask me. UCCUSS has a grainy look, i.e. there is high-frequency noise which is especially noticeable in darker areas, but no significant banding. The UCCUSS-1Selxo release however appears to have a poor translation with minimal typesetting, along with no chapters. Commie's translation is more important to the viewing experience than the slight improvement in video quality. Padpa encode is smoothed, reducing noise with same amount of banding, but with a slight reduction in detail as well. Some might prefer the clean noise-free look, but it is at best a trade-off here as it does not succeed in retaining details completely. Commie's encode has significant banding with slightly noticeable compression artefacts, at a detail level comparable to Padpa's release. The amount of compression artefacts are OK at that file size and not something you would spot without closer inspection and perfectly watchable, but the banding is easy to notice without making any comparisons against the source. Commie's encode does not qualify as 'BD quality' in my book because of the banding issues. TLDR: I would recommend Padpa's release, or mux and retime (if needed) Commie's translation to the UCCUSS encode if you can and are okay with a significantly larger file size.

>SMDegrain(out, tr=2, thSAD=100) Maybe that's why. Dunno anything about vaporsynth but I do know depending on source line art can be sensitive and easily destroyed (from experience) using that filter. I only use that on per scene basis. @Spiller The banding can be mitigated based on ones video player such as if using madVR's deband and dithering which would be recommended seeing it's h264 but can't bring back the washed out line art. Not sure why you'd recommend Padpa strictly on the term of video quality. No clue on how UCCUSS looks.

@Spiller UCCUSS is worse than Vardë's encode. It has unnecessary amount of grain and is linedarkened which ruins the lineart by giving it a dotted appearance. Slightly better detail retention means nothing when rest of the encode is filled with artifacts (like Commie as well). Vardë is obviously the best encode right now. [Comparison between Vardë and UCCUSS](https://slow.pics/c/ZdBq4VxU)