*** **TLDR at the bottom.** This is a passable release for 16:9 Bleach, with both Viz's original and modern script. Japanese sourced from DVDs (FLAC), English sourced from BDs (FLAC), sync by Handy. Remuxed using LogicalEgo's Rental ISOs. The mastering seems to be identical to STORE. *** Filtering: - **IVTC + hard 24p decimation;** Read below. - **Chroma repair;** Decombing YUV was insufficient for the interlacing issues these DVDs have, during my chroma denoise + AA pass I apply an RGB decomb. Decombing applied to luma too. - **Border repair;** Cropped to 718x480, vignetting border luma repaired, luma guided chroma warping 2px on each side. - **Edge repair;** Edge repair tackles line's mosquito noise. DeJPEGing was unfeasible given the scope of the project. Haloing present in these DVDs is bad to the point don't believe it can be 100% fixed. - **Denoising;** Temporal denoising, generally making the image static, then stable during movement. Dark-on-dark scenes are abominable, e.g. scuff marks on Shihakushous and on high frequency areas like distant rooftops. The denoising also acts as an impromptu deghoster, as these DVDs suffer from this abomination from studio deinterlacing. - **Crop + dither;** YUV420 causes noticeable loss, and my Crop values all require YUV444, I have encoded at YUV444P10. *** **IVTC explanation:** Afaict, TV broadcast NTSC DVDs produce with a 60i update rate, and all cinema (24fps), video (30fps) and broadcast (60fps) content is adapted to function in that container. Bleach was animated at 24p, but it was edited and mastered at 60i, resulting in a variable framerate (VFR) for different elements; sepia-grained flashbacks, scene transitions, and credits update at 60i (mostly all computer generated components are 60i), but so too do scene-boundaries. These "orphan fields" are solitary fields that lack the counterpart required for inverse telecine (IVTC) reconstruction. During editing, an original single frame becomes one field, and has no matching pair to reconstruct with. Recovery of these fields requires an IVTC tool that can accommodate a timescale of 1/60th of a second. Current automated solutions are capable of handling half-rate, but not these full-rate sequence(s). 24p-in-30p VFR has stutter, as opposed to judder if it was in 60p. All 60i content has been reduced to 24p but the underlying animation is rendered natively. Approximately 25% of scene-start and scene-end frames do not exist in any and all releases that do not deinterlace to 60p. One would need to deinterlace and splice these 60fps fragments into a VFR container. I developed a cadence differential (23.p vs 60p) script to detect these fields, allowing a spliced IVTC-60p video that reconstructs all fields, and interpolating these fields to 480p. This satisfies the requirement of 60fps TV while retaining IVTC's benefits. But the judder ends up being a bigger issue than the missing frames. A "global" automated solution for 198 episodes is unfeasible IMO. Field swapping can capture top/bottom orphan fields. This is why I chose 24p decimation. Globally reproducing the intended adence of the art itself is more important than the abilities of 90s era technology and bad modern support. I believe the SD production was inted, then deinted for 1080i mastering and storage (as in the underlying animation was deinted, then the 60i content was superimposed on top). Looking at the ITDL and ITBD (JP DVD masters) 1080i combing is visible, despite it having artifacts of SD deinterlacing. This same 1080i combing is apparent in the JPDVDs but as the vertical resolution is 4/9ths of the master, it is made redundant yet even on "progressive" frames, the same SD deinterlacing is visible. Bleach's interlacing is simply shit, there is no method I know of beyond manual IVTC that would achieve 100% restoration of the Master. Due to this, my IVTC settings are "conservative", airing on the side of leaving combing instead of crushing fields together. I tried a TAAU script, in an attempt to A. stabilize the SD line art, B. repair this SD deinterlacing and C. cleanup the residual combing. It is very expensive and doesn't respect the art: mouth flaps, eye- and character shaking are effected. The hallmarks of TAAU (ghosting, interpolation) are embedded into the content and "cannot" be fixed (e.g. mouth flap being a smooth 24fps when it was animated at 8). Due to this my decombing and denoising somewhat counters these artifacts, they are still visible but have been made more opaque. *** These DVDs are cropped (6 lines) compared to their SD 864x486i masters. To corroborate this, ITDL and ITBD (1080i masters) contain the full frame. Padding DVD 3px top and bottom matches these sources. Furthermore, production images also match IT's height. This release now matches the geometry of its original SD master, playing at 861x480. AR tags can only round to nearest pixel, this is a % of a % incorrect (861x480 instead of 861.6x480, 0.0697%). Production images show the pillarboxing in the masters did contain content, so their existence is not an element of squishing. The ratio goes from 15.95r:9.0 to 16.155:9.0, with pixel rounding, 15.956:9.0 to 16:144:9.0 (accommodating crop). I have experimented with 864x480 res, it has the added bonus of Crop(2,0,-2,0) to 860x480, but it's slow. I experimented with merging R2 and R1 DVDs together, it was moderately successful. It allowed for 720x483, as well as an averaged MPEG2 compression, But the difference in vertical resolution "forces" one to crop to 720x478. But the two sources are not the same, the masters are different thus the automated IVTC has discrepancies that causes desyncs. As the R2J DVDs are interlaced, they are the "best" source imo. The layers of abstraction between it and the raw production is the lowest. The ITDL has some notoriety as the best source available but its IVTC is bad, along with its decimation (despite being 29.p). An undecimated-IVTCed R2J has a sequence of A-B-B-C-D-E-F whereas ITDL has A-B-B-D-D-E-F sequences, dropping a C frame, and duplicating D. Decimating R2J displays as A-B-C-D-E-F, and for ITDL it ends up as A-B-D-D-E-F. This causes stutter and looks like shit, despite its higher resolution and 2020s H264 compression benefits. This same VFR headache is orders of magnitude more annoying with the deint-spliced IVTC, as I couldn't find a way to authentically decimate and de-dup at the same time. The IT sources are JP-DVD master based, neither version has good IVTC so stability is impossible unless you want to apply progressive repair to the image. I believe Bleach was produced at 960x540 and due to that, this release is inadequate on that basis alone, but the alternatives are worse or incomplete (as the native res is crushed). While these DVDs are the most authentic source, they are terrible. The SD deinterlacing is abominable, this is made especially clear in high motion - dark scenes, with distinct line-art. There is a good chance I have missed a core facet regarding DVD restoration - specifically aimed at tackling the issues I have detailed above. I have unfortunately then missed it. Subtitles were taken from HorribleSubs and EraiRaws batch 480p releases; the script resolution was resampled, omake announcements removed, Sushi was used to batch sync. Due to instances where Disney is delayed with its script, Crunchyroll is the default track. The Disney track contains English lyric subtitles and some script changes. ADDENDAM: Halfway through encoding I noticed Draknodd's encode, that maintains 60i content, marvelous work. Their v2 should be the gold standard Bleach encode. This also inspired me to revisit my filtering, so consider this a v1 of sorts. *** **Extras:** Regarding episodes 1-167; SOFCJ's is great for what it was, especially the effort of VFR in its earlier iterations. I have been working on a DVD rescale for it, and will see if I can find a way to keep 60i content (again) while incorporating the orphan fields. Those with JUMP page denoising experience, please also contact me via Discord, RGB SCANS. I attempted to add the title screen for each episode as their respective thumbnail. On Windows it doesn't appear but on mobile it does (sourced from Bleach wiki). I'm almost certain (some) animated segments were animated with 60fps viewing in mind, some segments with irregular frame-holds at 24p but at 60 the frame-holds become regular. 2019 iTunes is R1DVD master, 1280x720 @ 23. fps - deinterlacing issues. USBDs are 1080p, 168-224 is 23. fps, 225+ is 29. fps - major deinterlacing issues. JPBDs are a very warped, 1080i60 - deinterlacing issues. Crunchyroll was 1920x1080 for 266+ - deinterlacing and ghosting issues. Disney+ is USBD. PAL DVDs (R2FR) were lower bitrate than the R2J DVDs, technically speaking they probably "retain" Bleach's vertical production resolution. But as their bitrate is lower, and encodes a higher resolution, they are bad. Episodes ARs: 168-189 - 715x476 @ 854x476 (3, 4, -2, 0) 190-193 - 715x480 @ 858x480 (3, 0, -2, 0) 1.7875 194-201 - 718x480 @ 861x480 (1, 0, -1, 0) 1.7950 202-224 - 715x480 @ 858x480 (3, 0, -2, 0) 1.7875 225-366 - 718x480 @ 861x480 (1, 0, -1, 0) 1.7950 P.S. If someone could create an automated IVTC system that can accommodate 60i that would be great, if someone could create a YUV422 system that allows for full horizontal resolution that too would be great. *** **TLDR:** Japanese DVD, IVTC, decimated, denoised, cropped, dithered (16:9 718x480, 24p, YUV444, 10 bit, x264). Subtitles from both Crunchyroll and Disney. Italian sources have unfixable framerate issues. | Type | Name | Source | Format | Language | | :---: | :--- | :---: | :---: | :---: | Video | RENTAL DVD | R2J DVD | x264 | – | Audio 1 | DVD 2.0 | R2J DVD | FLAC | Japanese | Audio 2| BD 2.0 | USBD | FLAC | English | Subtitles 1 | Crunchyroll, Viz | Crunchyroll | ASS | English | Subtitles 2 | Disney, Viz | Disney+ | ASS | English | Chapters | R2J DVD | R2JDVD | XML | English | *** @ZeroviBritannia, R2J DVD, 540/576p is 50-70% more expensive to compute,