Koi Kaze

Public version of https://releases.moe/634/ [Join the Seadex discord server](https://discord.gg/rFwHDKHfJr) Video Track: 478p x264 10bit [WSE] Audio Track 1: Japanese 2.0 FLAC [WSE] Audio Track 2 (Default): English 2.0 AC3 [WSE] Subtitle Track 1: Full Subs [drawn-reality (edited)] Subtitle Track 2 (Default): Signs/Songs [drawn-reality (edited)] [Encode Comparison](https://slow.pics/c/C4lBYc9C) | [Subtitles](https://github.com/Wyse-/Koi-Kaze/) | [VapourSynth Scripts](https://github.com/Wyse-/anime-encodes/tree/master/Koi%20Kaze) | [x264 Encoding Script](https://github.com/Wyse-/anime-encodes/blob/master/Koi%20Kaze/encode_script/encode.sh) _EDIT: My encode has the wrong aspect ratio set. A v2 release with proper aspect ratio and some other improvements is in progress, this torrent will be deleted when it's ready._ This release has been in the works for quite a while, and is probably my biggest endeavour yet. I'll try to keep this brief, the existing encodes for Koi Kaze are quite bad: - GrimRipper used the inferior R1 DVD as source, and their encode's aspect ratio is way off - anime_fin used the R2J DVD as source, they maintained the correct aspect ratio, however their encode is completely destroyed by extremely heavy filtering Therefore, I made my own encode from the R2J DVD with the following goals: - Maximum detail preservation - IVTC done the best way possible, i.e. no combed frames and no automated decombing false positives - 29.970fps portions of the video handled properly, i.e. vfr encode - Correction of source issues, to the best of my ability The IVTC process was initially done automatically via [Wobbly](https://github.com/dubhater/Wobbly), however all subsequent combed frame fixing was handled manually either via Wobbly's GUI or later in the VapourSynth script. All 29.970fps parts (i.e. the scrolling text in the episode previews) were also properly handled via Wobbly, resulting in a vfr 23.976/29.970fps encode. As for correcting source issues: cropping was done in a way to preserve aspect ratio, interlaced fades were fixed via manually using [havsfunc's smartfademod](https://github.com/HomeOfVapourSynthEvolution/havsfunc/blob/master/havsfunc.py#L2008) and dirty lines were fixed via [rekt's rektlvls](https://gitlab.com/Ututu/rekt/-/blob/master/rekt/rektlvls.py#L64). In some cases, combed parts of frames were decombed manually through the use of [combmasks](https://github.com/Wyse-/anime-encodes/tree/master/Koi%20Kaze/combmasks) and [TDeintMod](https://github.com/HomeOfVapourSynthEvolution/VapourSynth-TDeintMod). Due to my lack of encoding experience and the goal to preserve detail, no other filters were applied to the source. Concerning the subtitles used: I heavily edited [drawn-reality's](https://github.com/drawn-reality/Koi-Kaze) subs, pretty much picking up where they left off in 2012 and added many improvements to styling and typesetting: - OP and ED karaoke styling redone from scratch using drawn-reality's translation - Dialogue track restyled - Typesetting redone from scratch for many signs - Typesetting slightly fixed for some signs, e.g. blur added, font changed, masks redone, etc - Added a few missing signs - Chidori's Judgement (i.e. the next ep previews) typesetting redone and properly typeset scrolling text signs added using R1 translation As for the rest, the motion tracking for signs was generally pretty good so I didn't have to redo any of it. Most of the dialogue is also untouched except two small edits for consistency and clarity: - [Original](https://github.com/drawn-reality/Koi-Kaze/blob/master/Koi%20Kaze%20-%2005.ass#L342) | [Edit](https://github.com/Wyse-/Koi-Kaze/blob/master/Koi%20Kaze%20-%2005.ass#L350) - [Original](https://github.com/drawn-reality/Koi-Kaze/blob/master/Koi%20Kaze%20-%2010.ass#L125) | [Edit](https://github.com/Wyse-/Koi-Kaze/blob/master/Koi%20Kaze%20-%2010.ass#L153) _I think that's it, for the most part. Enjoy!_