[Source-Previous_Best-Pizza Comparison](https://slow.pics/c/auHnM2eV) | [MediaInfo](https://pastebin.com/N85EpyYz) | [Vapoursynth Scripts](https://github.com/Moelancholy/Encode-Scripts/tree/master/Pizza/Kimagure%20Orange%20Road) Related: [OVAs] [Movie] [TV 720p](https://nyaa.si/view/1509969) (missing torrents will be uploaded at a later time) This v2 fixes minor issues with episodes 40 and 44, which made it impossible to skip around in the video (afaict the episodes still played back fine), because something or other went wrong during the transfer to the seedbox. --- **"Waaaah this is bloated!!"** FB-I = x265 2.4GB, Pizza 720p = 1.6GB, Pizza 1080p = 3.4GB. [I present this comparison without comment](https://slow.pics/c/UeqwVGES). The FB-I encode is decent for the most part and the only encode that even tried to fix the chroma while keeping the grain. I didn't like how it blurred/nuked detail in the reds and was significantly starved at times so I made this encode to watch the show myself. Ended up being a lot more work than expected for reasons that will become obvious in the following paragraphs. TL;DR: I fixed the chroma while keeping the encode otherwise transparent to source, though instead of completely removing *all* signs of chroma noise, I removed as much as could be removed without introducing significant issues. Since the main comparison was me taking random screenshots and discarding everything that wasn't a B frame, the main drawback of this encode isn't clear, so for the sake of transparency: this encode will harm specific dark blue gradients more than the FB-I encode, example [here](https://slow.pics/c/VtDWfSAJ). Fixing this would be trivial under normal circumstances, as scenes containing such gradients are fairly rare and could easily be scenefiltered. But I made this encode for myself to *then* watch it in the first place. Ideally, FB-I would have just linked their actual *Vapoursynth Script*, and not just the script they used to encode the already prefiltered lossless versions (???) so I could have just applied their method for that area of specific RGB values but alas. If anyone deems this a dealbreaker and wants to improve on my work, **all** my scripts are linked above. --- **Video** (Encoded from USBD because the JPBD was fucking useless) For the main episodes of the show, I cropped the video and applied a chroma denoise pass that was thresholded over the R plane to be weaker at extreme values to prevent blurring in the reds. Then I addressed the chromableed by using chromafix (partially warping the chroma, however this method is preferable to aWarpSharp) and applied some deringing. For ED1 the filtering was the same as the main episodes. For ED2 I did not apply any chroma denoising as it's an animation of coloured sand on glass which just gets nuked to high heavens otherwise (which would be a shame since it looks great). Instead I scenefiltered lumascaled deblocking where necessary. For ED3 I then again used the filtering from the main episode. All of the endings were zoned on a per-ending basis since they all needed different CRF from the main episode to be transparent. The openings were all spliced in from independently filtered lossless versions. For OP1, there didn't exist an NC version. I averaged together the openings from all the episodes featuring OP1 to mitigate ringing by linemasking it in (the complete averaged clip just leaves horrid looking dynamic blocking and no grain). I then applied chroma denoise with some scenefiltering, chromafix, deringing, and a scenefiltered lumascaled deblocking where necessary and zoned the opening in each episode. For OP2, it was a pain in the ass to filter because of the nature of the opening. I applied the same linemasked averaging, extensively scenefiltered chroma denoise and lumascaled deblocking, chromafix, and deringing. I zoned this opening in every episode. Also no NCs. OP3 had six versions. Yes, really. See [here](https://slow.pics/c/zxdbUQfi). Note that I didn't include the JPBD episode opening in this old comp because I knew it would just be worse. I ended up settling on the USBD NC, because it preserved the most detail apart from a few minor crushed blacks, had the largest part of the image, and the grain was the closest to the main episodes. Since it wasn't *actually* 4:3 I cropped it differently to this comparison later, to roughly respect the relative spacing of the openings found in the main episodes. I then recreated the fade in the beginning of the opening to last about as long as the fade-in found in the main episodes, and applied chroma denoise, chromafix and fairly strong deringing. As always I also zoned this opening in every episode it comes up. Comparisons: [OP1](https://slow.pics/c/4BRaK4lm) | [ED1](https://slow.pics/c/Uqlql3r7) | [OP2](https://slow.pics/c/sfilwN17) | [ED2](https://slow.pics/c/oTJ8iM9J) | [OP3](https://slow.pics/c/e3pD6RPm) | [ED3](https://slow.pics/c/kjlFmuCv) **Subs** PGS subs from the USBD. Same script as the softsubbed versions out there afaict (which likely are an old version of official subs), but with better punctuation and timing that does not make me want to commit vehicular manslaughter. If you can't cope with the subs being yellow use sub-gray. **Audio** Opus 2.0: --vbr --bitrate 192 Look I know some people are gonna say "lol lossy audio huge video" but I aim for transparency at the lowest bitrate possible and these audio bitrates are already far more than overkill. Stop coping. --- If there's major issues you can DM me on Discord @Moelancholy#6036. Also feel free to use any part of this torrent without permission or credit so long as you're actually doing something worthwhile. And use mpv.