Dragon Ball Z Project: R1 and R2J Dragon Boxes Merge

by JySzE and Romao∗

a Seed of Might Project
| | | |

January 22, 2021

This project continues Seed of Might’s trend of improving the raw footage from the
Dragon Box DVD sets due to the lack of quality official releases (see [5]) by creating an
innovative multi-sourced video track, getting the best out of both Dragon Ball Z Dragon
Box releases: the Japanese (Region 2 - 2003) and the American one (Region 1 - 2011).

Acknowledgements
For the video, big thanks to kratos and tm (theMaestro) for first bringing to light the
overall better dark details in R1 Dragon Box. Without their insight, this project probably would
have not even been done.
For the audio, we thank kei17 (and whoever helped him) for collecting the various broadcast
audios, AnimeMaakuo and HKB for releasing them, Team Mirolo for carefully synchroniz-
ing all the audio files to R2J Dragon Box video, and iKaos for adding the Dragon Ball GT’s
first episode preview to the end of the final episode.
For the subtitles, we thank many people. Marin888 for cleaning-up and converting Romao’s
OCR of R1’s VOD subtitles to SRT with proper line breaks. Team Mirolo for further synchro-
nizing them to video. Evoshare for the translation of Japanese signs during episodes. corre
for formatting them with amazing custom titlecards and signs. iKaos for quality-checking and
improving their formatting and styling. dubadavid for the opening credits. SaiyaJedi for song
translations. Finally, dubadavid, AnimeAjay, and many others for helping with translations
and the update on name adaptations.
For the chapters, we thank gravitypriest for adding an extra chapter to the titlecards.
We thank tm and iKaos again for reviewing and improving this very document.

1 Tracks
The Matroska file of each episode contains the following data:
Track Type Name Format Language
0 Video Dragon Ball Z [SoM] AVC —
1 Audio 1989 Broadcast Audio [TeamMirolo] FLAC Japanese
2 Subtitles Styled Subtitles [SoM] ASS English
3 Chapters — text/XML English

Attachment Type Name Format

0 cover — JPEG
1 font Funimitation TTF
2 font Faxfont Tone OTF
3 font Oravetica TTF
4,5 font Josephs Brush (Regular, Italic) TTF
6 font Wolf Sans TTF
7 font Caballar TTF
8 font Tomorrow Comes Today TTF
9 font PP Handwriting TTF
∗
a.k.a. John Rambo in Discord and other media platforms

1
2 Video
This is not a purist project. For those seeking this kind of release, there are already remuxes
and encoded batches with less modifications to the image available for download ([4],[2]).
Purism is collectively understood as “you don’t fuck it up if you don’t touch it.” It is an
important and safe attitude that keeps decent sources from disappearing and helps people get
more critical of poorly made edits and filterings.
However, purism for the sake of purism can sometimes be limiting. With that in mind,
the intention of this release was to improve upon the Dragon Boxes by not only reducing or
eliminating flaws in the source material, but also combining the strengths of each release together
in a attempt to make the video look more like the original Dragon Box masters. Rest assured
that this will not be some blue-tinted warpsharped mess, but rather a pretty based release that
is quite unlike anything you have ever seen before.
Criticism is welcome, by the way!

2.1 Metadata
The video tracks of the episodes have the following properties.

Codec ID FR1 DAR2 PAR3 Resolution Bit depth Planar format

V MPEG4/ constant
4:3 160:177 708 × 480 8 bits YUV420
ISO/AVC 24000/1001

Encoding settings
cabac 1 ref 6 deblock 1:-3:-3
analyse 0x3:0x133 me umh subme 10
psy 1 psy rd 1.00:0.25 mixed ref 1
me range 32 chroma me 1 trellis 2
8x8dct 1 cqm 0 deadzone 6,6
fast pskip 1 chroma qp offset -4 threads 16
lookahead threads 3 sliced threads 0 nr 0
decimate 0 interlaced 0 bluray compat 0
constrained intra 0 bframes 16 b pyramid 2
b adapt 2 b bias 0 direct 3
weightb 1 open gop 0 weightp 2
keyint 250 keyint min 23 scenecut 40
intra refresh 0 rc crf mbtree 0
crf 13.0 qcomp 0.80 qpmin 0
qpmax 69 qpstep 4 ip ratio 1.10
pb ratio 1.10 aq 3:0.50 library x264

Both R1 and R2J are 720 x 480 resolution, which you might think looks horizontally
stretched, but they come with a flag that instructs the intended display to resize it to how
it should actually look. This flag is something called a pixel aspect ratio, or PAR. It can be
confusing to wrap your head around, but the basic idea is that it tells the display that each pixel
in the 720 x 480 resolution should be taller than it is wide by a set amount. This will accurately
resize the image and make it look as it should. This PAR transformation plus overscan from a
CRT television (the intended display) yields a 4:3 picture (the display aspect ratio, or “DAR”).
1
Frame Rate (frames per second)
2
Display Aspect Ratio
3
Pixel Aspect Ratio

2
Today though, we don’t use CRTs and so some decisions have to be made about how to resize
accurately. One thing is clear: it is not possible to have a 4:3 DAR and the correct PAR without
cropping. So we are left with 3 options.

1. Prioritize a 4:3 DAR at all costs without cropping. This leads to an incorrect PAR and so
the image will look vertically stretched compared to what is intended.

2. Prioritize the correct PAR without cropping. This leads to a picture that looks correct
but is wider than 4:3 and will sometimes show part of the image that is not meant to be
shown (e.g. inconsistent and blurred black bars), since DVD authoring always predicts
overscan by the old CRTs.

3. Prioritize the correct PAR and a 4:3 DAR, which is achieved by cropping away the over-
scanned part. This results in resolutions such as 708 x 480 (which is resized to 708 x 531
on most players after the PAR transformation).

This project uses the third option.

2.2 Pixel aspect ratio

Non-square pixels were supposedly a smart choice at the time. The standard was analog
display that could pack more information horizontally than vertically given any square sample of
it. The industry had to optimize the data on DVDs for the consumer end, so this mathematical
pixel concept that can only be approximated by modern players made sense in the past.
If you search right now for PAR correction on DVD footage, you will probably get to the ratio
10:11, both on simple and very detailed explanations on CRT analog-to-digital, signal-to-pixel
conversions. But this is not the ratio we are using (160:177).
Another wrong PAR is 8:9, which comes from the resolution standard first set to DVDs
fitted on a 4:3 DAR, which has nothing to do with how CRTs were using the converted signal
to display a 4:3 footage. A couple of standards were written to fix the squeezed DVD footage.
These standards were digital standards, to match the digital display of the video to what a
well-calibrated CRT already would correctly display. Thus it is a conversion standard. From
8:9 to whatever gives the correct image. It is a complicated matter and the numbers get pretty
ugly. Even the nice recommended 10:11 ratio is actually an approximation to 42651:38800 [1].
The 160:177 PAR standard comes from SMPTE RP 187-1995 [6] and supposedly solved the
rectangle pixel issue for good. The industry ignores it nonetheless, because the numbers they
came up with were awfully obnoxious and even the scaling process would be more hardware-
expensive. This explains why this standard is not as well-known as the 10:11 one. But here in
2021, scaling video is not a problem, and your player will scale the footage like it is nothing.
We have tested it, and 160:177 looks best overall on Dragon Box. Meticulous comparisons
using circle objects in the video (like the dragon ball at the very beginning of each episode)
didn’t give a huge edge to 160:177 over 10:11 because the film itself always has some form of
warping. However, it is also better to be safe than sorry. Having a 4:3, 160:177 PAR footage
leaves it with 4 extra pixel columns compared to a 4:3 with 10:11 PAR. You can still crop them
away and set the PAR to 10:11 if you want, which is not possible to do the other way around.
We refer to the lurker’s guide to video for more detail on pixel aspect ratio [3].

2.3 About the merging of the R1 and R2J footage

This merge of the two Dragon Boxes aims to get more detail at the boundary ranges of
luma in order to obtain a higher dynamic range in a natural manner, without artificially warp-
ing/crushing the luma using sigmoid-esque functions. This is made possible due to the brightness
difference between the North American NTSC and the Japanese NTSC-J standards and how
the DVD compression codec (MPEG-2) handles this difference. Basically, MPEG-2 gives more

3
bitrate to brighter areas. Darker areas get less bitrate and so the image details there are blurrier
and often destroyed. Fortunately for us, the North American NTSC standard has brighter blacks
compared to NTSC-J, which means that MPEG-2 was able to allocate more bitrate to the dark
areas on the R1 Dragon Box compared to the R2J, even though the latter has a higher overall
bitrate. In addition to better dark details, the R1 Dragon Box also has more dark details. This
is because DVDs have a limited luma range, and the brighter blacks on the R1 allowed more
dark details to pass through that limited range. These same extra details missed the cut on R2J
and were clipped away instead. So what does all this mean? It means that the R1 Dragon Box
has better preserved dark details while the R2J Dragon Box has better preserved mid-and-bright
details.
And now with the merge, we can combine the best of both worlds in order to obtain an
image quality that is superior to either source on its own. We believe that this merged release is
a much closer representation of what the Dragon Box masters look like. Get ready to experience
a level of detail and dynamic range that has not been seen before from the Dragon Box.
One thing to note is that the detail difference between R1 and R2J can vary. This means
that while some shots will show a profound improvement in the merge, others may only reveal
a subtle difference. The figure below exemplifies the general differences between R1 and R2J
with their mid tones matched. R2J is clearly sharper compared to R1, yet R1 has contours in
dark areas which are not as evident in R2J.

R1 R2J

2.4 Blocking treatment

R2J has a lot of compression in its grain in the form of flat macroblocks with constant luma
values, inside of which there is no relevant information whatsoever. In digital footage with little
noise, smoothing away macroblocks blends well with the rest of the footage, but that is not the
case with footage filled with film grain.
As a better approach to this type of video image, we filled these compressed areas with grain
based on the grain of the footage to make it more uniform and pleasant to watch, again bringing
the watching experience closer to what watching the SD masters would be. For the areas where
there is already grain (with no significant compression) and detail, the image left untouched,
thus preserving every bit of detail of the raw merged video.
Another important thing we did is to restrict the strength of the grain applied on the flat
blocks in order to ensure that we are not visually destroying the existing detail. This prevents
subtle details from being less perceptible when neighboured by noise with high variance. The
result is a natural-looking, grain-preserving compression cleaning.

4
 R2J Ours

2.5 Chroma treatment

Sadly, we cannot give chroma the same noise treatment. The luma noise has a certain amount
of detail allowing it to be kept and enhanced. The chroma channels, on the other hand, are
half the resolution of luma and have far more compression. What would otherwise be fine color
noise from the film is just broken phagocytized blobs of compression that only help in making
the footage unappealing. They had to go.

R2J Ours
Regular Display
Noise Enhanced

Another major problem with the video source is chromatic aberration. It is pretty strong in
some episodes. Regular filters usually do not work well in this case and most of them have
unacceptable drawbacks. Our solution will clean it almost completely, but sometimes soften
the chroma channels where chroma aberration is detected. We made sure to mitigate this issue
by further improving condition masks and calibrating the strength of our script to balance the
softness and aberration to our liking.

5
 R2J Ours

Finally, one might ask if there were changes in colors or any form of color-correction. We did
not do any sort of color-correction for this release. Some of the episodes will look less tinted on
this release compared to R2J, but there are two main reasons for this. Firstly, the clipped dark
tones in R2J makes the dark tint relatively more visible even with the same chroma channels.
Secondly, some R1 episodes just randomly have less of a tint, perhaps because Toei Animation
messed up those episodes for the R2J, and chroma from R1 is used in this project. Anyone
developing a scene-by-scene white-balance project for R2J should not worry though, since R1’s
ocasionnal lower tint is probably on the whole episode or chapter. So R2J white-balance applied
onto this project’s video should give an image with constant amount of tint, which is easy to fix.

3 Audio
If you still do not know what the audio of the original broadcast of Dragon Ball Z is all
about, just search for it, honestly.
The following table discriminates the audio source of the episodes. Tokai, Ishikawa and Fuji
are the Japanese television channels through which the audio was first transmitted.

Source Episodes
009, 017, 019, 032-045, 100, 105, 107, 109, 114-141,
Tokai
143-191, 198-211, 213-216, 218-225, 227-291
006-008, 010-013, 015-016, 018, 020-030, 046-050, 052-057,
Ishikawa
059-061, 064-094, 096-101, 104, 106, 108, 110-114, 212, 217, 226
Fuji 001, 003-005, 014, 031, 051, 058, 062-063, 095, 142, 192-197
Drama Cassette 002
Big Box CD 229, 237

Episodes 100 and 114 are both present in the Tokai and Ishikawa row, which indicates they are
composed of both sources. There are also “next episode previews” (NEPs) missing from some
of these broadcast audios. Team Mirolo used a filtered version of the Dragon Box audio to fill
in these missing NEPs. These episodes are: 009, 037-039, 052-054, 079-081, 097-099, 128-130,
145-147, 175-176, 190-192, 218-220, 230-232, 256-258, 268-272, and 290.

3.1 Metadata
The single audio track of each episodes have the following properties.

Codec ID Sampling rate Bit depth Compression Channel(s) Library

A FLAC 48 kHz 16 bits Lossless 1 libFLAC

6
3.2 Team Mirolo’s sync
From their own words:

The Japanese Dragon Box Z DVDs have their audio tracks badly synced. Sound
effects appear later than they should. On average, sound effects appear 200 ms too
late.
In order to video-sync, we had to calculate the gap in milliseconds between many
video frames and their respective sound effects, for both part one and part two of
each episode.

Toei Animation sync

SOUND EFFECT

Team Mirolo sync

SOUND EFFECT

The set of tables at the end of this document shows the difference in milliseconds to get
from the original dragon box audio (or any audio track in sync with it) to the video-sync. In
that table, “Part A” refers to the first half of the episode, from the beginning to the end of the
first eyecatch. “Part B” refers to the second half, which is everything after Part A. The “Diff”
is simply Part B minus Part A, which is what needs to be change in Part B in case the fix for
Part A is already applied to the entire episode.
Team Mirolo attaches an updated table with these time shifts to most of their releases.
It is present in this document for the purpose of making it easy for people to update the
synchronization in case Team Mirolo updates their sync.

4 Subtitles
As mentioned in the Acknowledgements section, subtitles in the opening, ending, and title
cards are matching the Japanese kanji present in the Dragon Boxes.
Our subtitles are based on Steven J. Simmons’ translation with some Herms reviews taken
into account. These subtitles were missing translations of Japanese text displayed in the middle
of the episodes. The episodes for which new subtitles were added to untranslated Japanese signs
or other forms of text are the folowing: 013, 030, 049, 061, 089, 137, 144, 171-172, 186, 195-197,
208-209, 214-215, 218, 223, 243, 249, and 289-291
The changes in the names and special words took four major factors: consistency, adaptation,
modern and updated romaji, and language origin of the word or the pun indended for the word.

7
 Simmons Ours
Anoyoichi Tournament Anoyoichi Budokai
Artificial Human Cyborg
Chaozu Chiaotzu
Demon Clan Mazoku
Demon Clansman member of the Mazoku
Demon Clansmen Mazoku soldier
Ghurd Gurd
Great King Yama Enma Daio
Gyallic Ho Gallick Cannon
Gyumao Gyumao
Kamehameha/Kamehame-Ha Kamehameha
Kami-sama/Kami God
katchinko (metal) katchin (steel)
Kiai-Ho Kiaiho
Kibit Kibito
Kikouhou Kikoho
Kiko–Ha Kikoha
kili kiri
Kinto Un Kinto’Un
Madoushi Warlock
Ma-Junior Ma Junior
Nyoibou Nyoi-bo
Pafu pafu
Polunga Porunga
Pu-erh Pu’er
ReaCoom Reacoom
senzu (bean/s) senzu
Shenlong Shen Long
Shishin no ken Shishin no Ken
Shu Saiaku Shu Sai’aku
Taiyou-ken Taiyo-ken
Tao Pai Pai Tao Pai-Pai
Tenkaichi Tournament Tenkaichi Budokai
The Makyo World Planet Makyo
Yama Enma
Yamucha Yamcha

References
[1] Die bedeutung der itu-r bt.601 für das par. https://encodingwissen.de/hintergrund/
videobild/anamorph/itu-r-bt601/. Accessed: 2020-12-18.
[2] ikaos multi-audio releases on nyaa. https://nyaa.si/?f=0&c=1_0&q=iKaos. Accessed:
2020-12-18.
[3] The lurker’s guide to video. https://lurkertech.com/lg/video-systems/#pixelaspect.
Accessed: 2020-12-18.
[4] Romao’s dragon box z remux. https://nyaa.si/view/1114198. Accessed: 2020-12-18.
[5] Seed of might releases on nyaa. https://nyaa.si/?f=0&c=0_0&q=%22%5BSoM%5D%22. Ac-
cessed: 2010-12-18.
[6] SMPTE, RP. Rp 187-1995. Center, Aspect Ratio and Blanking of Video Images. doi:
10.5594/SMPTE.RP187.1995.

8
Episode Part A Part B Diff Episode Part A Part B Diff
001 -110 -90 20 051 -200 -230 -30
002 -150 -150 0 052 -200 -250 -50
003 -150 -150 0 053 -210 -250 -40
004 -120 -120 0 054 -110 -180 -70
005 -180 -150 30 055 -200 -200 0
006 -140 -140 0 056 -220 -260 -40
007 -150 -150 0 057 -120 -190 -70
008 -110 -110 0 058 -200 -140 60
009 -130 -130 0 059 -170 -220 -50
010 -150 -180 -30 060 -110 -90 20
011 -160 -160 0 061 -140 -160 -20
012 -110 -110 0 062 -180 -180 0
013 -130 -200 -70 063 -190 -150 40
014 -220 -160 60 064 -170 -180 -10
015 -180 -150 30 065 -110 -290 -180
016 -80 -80 0 066 -150 -150 0
017 -100 -100 0 067 -170 -250 -80
018 -210 -150 60 068 -240 -90 150
019 -120 -180 -60 069 -170 -160 10
020 -200 -120 80 070 -190 -150 40
021 -120 -120 0 071 -200 -230 -30
022 -110 -110 0 072 -200 -180 20
023 -50 -150 -100 073 -180 -30 150
024 -150 -180 -30 074 -210 -200 10
025 -180 -200 -20 075 -210 -210 0
026 -180 -180 0 076 -130 -230 -100
027 -180 -200 -20 077 -190 -170 20
028 -130 -180 -50 078 -170 -170 0
029 -80 -120 -40 079 -200 -220 -20
030 -160 -160 0 080 -140 -200 -60
031 -170 0 170 081 -160 -160 0
032 -170 -170 0 082 -180 -210 -30
033 -150 -150 0 083 -240 -160 80
034 -150 -150 0 084 -90 -160 -70
035 -150 -150 0 085 -210 -210 0
036 -210 -150 60 086 -230 -230 0
037 -230 -210 20 087 -180 -180 0
038 -250 -180 70 088 -170 -230 -60
039 -220 -100 120 089 -180 -190 -10
040 -120 -220 -100 090 -160 -90 70
041 -190 -220 -30 091 -210 -210 0
042 -180 -210 -30 092 -250 -250 0
043 -180 -160 20 093 -200 -200 0
044 -190 -190 0 094 -230 -180 50
045 -270 -270 0 095 0 -170 -170
046 -210 -230 -20 096 -180 -210 -30
047 -210 -230 -20 097 -230 -140 90
048 -150 -210 -60 098 -160 -320 -160
049 -230 -230 0 099 -270 -140 130
050 -200 -210 -10 100 -70 -100 -30
9
Episode Part A Part B Diff Episode Part A Part B Diff
101 -220 -130 90 151 -170 -200 -30
102 -200 -200 0 152 -190 -190 0
103 -210 -230 -20 153 -240 -180 60
104 -260 -240 20 154 -240 -190 50
105 -260 -260 0 155 -220 -260 -40
106 -220 -220 0 156 -230 -230 0
107 -190 -190 0 157 -200 -200 0
108 -210 -260 -50 158 -190 -190 0
109 -240 -240 0 159 -230 -180 50
110 -230 -180 50 160 -210 -210 0
111 -160 -260 -100 161 -220 -220 0
112 -200 -240 -40 162 -210 -260 -50
113 -180 -200 -20 163 -220 -250 -30
114 -190 -220 -30 164 -240 -180 60
115 -200 -200 0 165 -260 -260 0
116 -200 -200 0 166 -230 -150 80
117 -190 -190 0 167 -180 -190 -10
118 -250 -250 0 168 -220 -180 40
119 -220 -240 -20 169 -240 -130 110
120 -230 -230 0 170 -200 -160 40
121 -200 -190 10 171 -250 -200 50
122 -220 -210 10 172 -240 -150 90
123 -230 -250 -20 173 -240 -130 110
124 -180 -180 0 174 -230 -210 20
125 -170 -170 0 175 -190 -210 -20
126 -230 -180 50 176 -190 -210 -20
127 -150 -180 -30 177 -180 -180 0
128 -220 -220 0 178 -200 -180 20
129 -180 -240 -60 179 -240 -200 40
130 -140 -210 -70 180 -230 -170 60
131 -180 -200 -20 181 -220 -230 -10
132 -210 -200 10 182 -200 -200 0
133 -230 -230 0 183 -230 -260 -30
134 -200 -250 -50 184 -180 -120 60
135 -190 -150 40 185 -180 -220 -40
136 -180 -180 0 186 -220 -160 60
137 0 -150 -150 187 -190 -200 -10
138 -160 -240 -80 188 -220 -160 60
139 -180 -180 0 189 -210 -170 40
140 -200 -170 30 190 -200 -160 40
141 -260 -260 0 191 -170 -130 40
142 -150 -230 -80 192 -240 -200 40
143 -190 -190 0 193 -170 -210 -40
144 -230 -200 30 194 -220 -150 70
145 -150 -180 -30 195 -210 -170 40
146 -230 -230 0 196 -210 -180 30
147 -190 -230 -40 197 -190 -170 20
148 -250 -250 0 198 -260 -240 20
149 -170 -200 -30 199 -200 -130 70
150 -160 -200 -40 200 -240 -100 140
10
Episode Part A Part B Diff Episode Part A Part B Diff
201 -140 -150 -10 251 -240 -180 60
202 -200 -160 40 252 -220 -180 40
203 -170 -150 20 253 -180 -160 20
204 -190 -150 40 254 -160 -150 10
205 -190 -170 20 255 -240 -170 70
206 -190 -210 -20 256 -190 -150 40
207 -200 -160 40 257 -220 -190 30
208 -180 -180 0 258 -260 -200 60
209 -250 -250 0 259 -250 -170 80
210 -210 -250 -40 260 -150 -170 -20
211 -180 -250 -70 261 -190 -190 0
212 -190 -220 -30 262 -210 -210 0
213 -210 -230 -20 263 -230 -190 40
214 -180 -200 -20 264 -210 -170 40
215 -190 -170 20 265 -190 -190 0
216 -190 -190 0 266 -250 -190 60
217 -210 -170 40 267 -250 -140 110
218 -140 -100 40 268 -230 -190 40
219 -210 -170 40 269 -250 -210 40
220 -110 -220 -110 270 -200 -190 10
221 -180 -250 -70 271 -180 -190 -10
222 -190 -190 0 272 -230 -230 0
223 -150 -130 20 273 -200 -180 20
224 -190 -210 -20 274 -250 -180 70
225 -200 -180 20 275 -240 -200 40
226 -200 -180 20 276 -240 -220 20
227 -180 -210 -30 277 -220 -230 -10
228 -200 -200 0 278 -220 -180 40
229 -130 -200 -70 279 -240 -200 40
230 -180 -200 -20 280 -230 -190 40
231 -210 -210 0 281 -250 -210 40
232 -180 -220 -40 282 -220 -240 -20
233 -200 -200 0 283 -220 -180 40
234 -210 -180 30 284 -240 -170 70
235 -170 -210 -40 285 -180 -250 -70
236 -180 -180 0 286 -260 -270 -10
237 -260 -260 0 287 -200 -200 0
238 -190 -190 0 288 -210 -170 40
239 -190 -170 20 289 -210 -170 40
240 -90 -50 40 290 -240 -170 70
241 -160 -120 40 291 -240 -200 40
242 -200 -160 40
243 -220 -180 40
244 -180 -140 40
245 -200 -170 30
246 -200 -100 100
247 -200 -140 60
248 -190 -150 40
249 -210 -170 40
250 -150 -190 -40
11