Reporting MTL releases

Posted in

#151 by eacil
2021-09-18 at 10:05
< report >

Open source translation for the Tsukihime Remake game prepared by Clovermoon and Tsukihimates.

Welcome to the repository of the Tsukihime Remake open source translation. It contains all the resources of the game (text, pictures and videos) that can and should be translated. Here is a general spreadsheet containing all the information about the advancement of the translation for all of those abovementioned resources: link

This project will use heavily the deepLuna extraction/edition/translation/injection script that is available in the following repository: link

The repository is currently still under building.

deepLuna, from "deepL", the machine translation service, and "Luna", the name of the Roman Moon goddess, is a Python script with a GUI interface that works as an extractor/editor/translator (with deepL API)/injector for the text of the Tsukihime Remake game on Switch and possibly on PS4, while taking into account the whole internal structure of the script.

So, unless you show some actual sources instead of doing your usual half assed job of reporting random things, I am not reverting the MTL flag.
Editing this MTL with "real translators" will not change this decision and you know it.
#152 by lemonov
2021-09-18 at 18:04
< report >#151 link
They just claimed that they'll do it, and they are doing it already, accorind to their spreadsheet that has plenty of non-machine JP=>EN colored lines. Sure, it uses MTL as a base that a real translator corrects, but that still counts since there is the real translator involved. We have plenty of translations which were clearly based on other translations on this website already.
If you don't concider this a real translation effort, then you might as well remove r76760 because they haven't even started working on it (and likely never will).
#153 by hansfranz77
2021-09-18 at 20:27
< report >
Editing this MTL with "real translators" will not change this decision and you know it.
I wonder when we will get folks screaming about "edited mtl vs mtl", because at the end of the day both still are mtl, no matter how much you want to convince yourself otherwise.

Btw that is also why i didn't respond to you in the fate re-tl thread anymore. There seems to be no argument which gets accepted with folks defending edited mtl, which is just sad.Last modified on 2021-09-18 at 20:29
#154 by eacil
2021-09-18 at 21:12
< report >Well, this is not this easy of a subject because you have two types of edited MTL.

If edited MTL is frown upon, this is because, in most cases, the editor knows jack shit about Japanese. Meaning that neither him or the algorithm have a single grasp on the underlying script. You could argue that a translator could be so bad that he doesn't have a better grasp on the script than a MTL and you would be right, but we are not judging quality and draw the line at, atm, human translator > MTL, basically considering that the MTL is pretty much a blind translator (for reason as simple as having no good grasp on the context).

However, if MTL is used as a base and a team of translators are reviewing every line, this is wholly different. The MTL is used to speed things up. We are not talking about f95zone level of garbage, but something worth considering. I know DeepL is better than me at writing English sentences but then I am not a good example. At least, the accuracy should be here and that's pretty much all that matters in this debate. To sum things up, you are skipping the TL part and starting directly with TLC. The TLC step can be as long as the actual TL step depending on who does it.
The problem lies on how lazy the translator team is. With MTL, the temptation of superficially TLCing is big. You don't have to do anything. Without MTL, you can't just do nothing, the lines won't be written all by themselves.

You have the middle ground with edited MTL and vague QC using the Japanese. This is the latest scam with your average Joe and their ten years of watching anime experience looking at the script and... I don't know... checking if the names matches? If the MTL didn't fuck itadakimasu up? This is especially popular with folks using a pre-existing translation. They can't translate using Japanese but, fear not, they QCed using the original script. As if you could trust this level of absolute bullshit. At best, you will have the most basic Japanese sentences being checked. Pretty much what everybody, even MTL, have no problem with most of the time.
How many people pretend to have a "basic understanding of Japanese" when, in fine, they suck balls at it?

Anyway, not all edited MTL are equal and technically you could consider one being a "good" translation, but this depends on how much you can trust the team in the first place. I am not sure I would trust anybody imo. If you are serious and good, you don't use MTL as a base, MTL as a tool to help you when you are stuck, sure.

We have plenty of translations which were clearly based on other translations on this website already.
It has nothing to do with MTL because, contrary to MTL, the base is considered trustworthy, not being made by "blind" translators. Also, if another language was chosen, it's because the translator is "blind" to Japanese, and the reason why they are using a proxy language. The problem lies elsewhere as it's about how much you lose from translating a translation.
#155 by lemonov
2021-09-19 at 01:43
< report >From where I stand, r82663 should remain being marked as MTL until the team actually delivers on "translators are reviewing every line" part. For now, sure, keep it as is.
#156 by cubky
2021-09-19 at 05:19
< report >Editted MTL is still MTL. Does the quality of the result change if the TLC is done by a Japanese translator, average Joe or a trained monkey? Very likely yes, but that does not change what the translation should be marked as.

Nobody is saying anything about the quality of the translation by using the MTL tag. If you think they do, then you should probably think about why that might be.Last modified on 2021-09-19 at 05:22
#157 by sonicokyu
2021-09-19 at 06:28
< report >More evidence of r54565 being an MTL





Bunch of weirdly untranslated bits and things that were already mentioned before. It's still a MTL regardless of how they want to deny. Why it hasn't been tagged is weird.
#158 by naiohoras
2021-10-02 at 08:10
< report >r83564.1 r73761.2 r83563.1 r74860.4 r83562.1 r83561.2 r83561.1 r83560.1 r74859.2 r76973.2 r75866.2 r83558.1 r76994.2 r75865.3 r77574.2 r83557.1 r83556.1 r83555.1 r83554.1 r73762.2 r83553.1 r75864.2 r75863.3 r72178.4 r73763.2 r76993.3 r83552.1 r83551.1 r83550.1 r73759.2 r83549.2 r83548.2 r75867.4
#159 by hirary
2021-10-03 at 05:33
< report >This is obv MTL. They even say it in a phew releases
#160 by poudink
2021-10-03 at 05:34
< report >The description says only some of their translations are MTL. Is there any proof of the contrary?
#161 by tomyun
2021-10-03 at 05:55
< report >So I marked them down as 'MTL' in the note when they're based on MTL. You can distinguish them by seeing "간이" (簡易, literally meaning "simple") in their patch titles. For the rest, they're indeed hand translations, often accompanied with images and movies edited as well. Nowadays you can view this team as an almost entire Korean VN translation community integrated.
#162 by eacil
2021-10-03 at 06:02
< report >A bazillion releases in two years. I say I will all mark them as MTL cause I have something else to do than waste my time checking all that crap with the hope than one or two are not.
If one is not, make a case.

Also, what's the difference between r82675 and r83576?

Edit: I guess I will follow your lead tomyun.Last modified on 2021-10-03 at 06:19
#163 by tomyun
2021-10-03 at 06:08
< report >One thing I wanted to add is that they don't actively advertise themselves under the name "Team Myskrpatch" most of the time. It's more like independent projects happen to be distributed from its website while a handful extremely competent core members helping these projects. Usually there is no overlap in translators and that's one reason how they were able to keep up with so many quality patches in a relatively short time frame. When there is no dedicated translator, that's when you see these MTL releases.
#164 by tomyun
2021-10-03 at 06:20
< report >For Tsukihime, they first released the MTL patch (r82675) last month as a technical proof of concept while they also teamed up with a dedicated translator to complete hand translation day and night. Then the complete patch (r83576) just released yesterday. There's a post mortem posted (in Korean) by the said translator if anyone interested.Last modified on 2021-10-03 at 06:22
#165 by eacil
2021-10-03 at 06:27
< report >Another one who worked 16 hours a day for two months? Ah, no, contrary to r82202, he completed a longer VN in half the time. Man, those translators are faster and faster every day.
#166 by tomyun
2021-10-03 at 06:47
< report >It's not uncommon for Japanese-Korean translation taking much less effort compared with other languages, but still he was like a beast at least in terms of speed. He was actually quite well known for his past projects like r70745, r74857, r82210 where he seemingly worked alone as well. Anyways, Korean community is now very excited with long-awaited high profile games (Tsukihime, White Album 2) just dropped in a day apart and I haven't seen anyone complaining the quality of translation so far.Last modified on 2021-10-03 at 07:08
#167 by tomyun
2021-10-04 at 05:41
< report >some slipped out of mass editing: r73759 r74859Last modified on 2021-10-04 at 05:42
#168 by 4digitmen
2021-10-06 at 01:21
< report >r80821 according to the description and the repository linked within link
#169 by eacil
2021-10-06 at 09:52
< report >We decided to not mark as MTL releases which have a minor fraction of MTL in them, like it seems to be the case here. The rationale is that even if you are against MTL, you might want to know that a partially legit translation exists. It's pretty much the equivalent of a partial release for an anti-MTL. In this case, you can play almost all of Higurashi on Switch and there is no reason to hide that.
So far, it only concerns ports which fill the holes with MTL.

Edit: imo, only release will well identified MTL parts should not be marked as MTL. If it contains an indefinite amount of MTL on indefinite places, it should be marked as MTL because we can't tell what to avoid.Last modified on 2021-10-06 at 09:55
#170 by 4digitmen
2021-10-06 at 13:43
< report >I'd hardly call it minor given all the Hou Arcs and more are 100% machine-translated, but sure, I guess that's fine.
#171 by cubky
2021-10-06 at 13:58
< report >I feel like if the release can stay not tagged as MTL as whole, then it should have the description saying which parts are MTL. Currently it has the vague "Using partially machine translation in some chapters" and you have to go through the VN entries to try and guess which parts might have proper translation.

Edit: Now I noticed the description is instead placed in the github repository, so I guess that could be enough.Last modified on 2021-10-06 at 14:08
#172 by eacil
2021-10-06 at 14:22
< report >#170: I didn't mean minor in the sense of negligible but in regard to the whole content. It doesn't really matter in fact because I am sure some people will be happy to know that they can play the entirety of Higurashi no Naku Koro ni on Switch and for that only it's worth not tagging the release as MTL I believe.

#171: I agree, that's why I edited my post. If you know which parts are MTLed, feel free to edit the release, even if you copy-paste stuff from the outside.
#173 by kevinbad2010
2021-10-09 at 20:42
< report >r83676 is an MTL patch
#174 by ninigi
2021-10-18 at 17:14
< report >link

#175 by ernovace
2021-10-23 at 13:30
< report >link

From the website linked for the patch: link

> Language: English (MTL and heavily edited)