raw: /g/2976204/82c7c9a5cc/

Base +6, xXDGFXX +6, Aside +6, Meccaflare +8, MehMeher +9

Go fuck yourself

Base +11, faytT +8, badgerkins +7, kafi +15, barea100 -6, Addelheim +24, SuperGenericAnon +8, Dante456 +11, Higers +7, deltski +6, IngmarS +8, RandomDudeH +6, pesc +8, and 33 more...

torrent link download pls!

Base +6, Higers -7, IngmarS -8, as102 -13, fap222222 -6, Undeadcynic97 -9, LillySatou -7, itsaferbie -8, lord-max -13, dankestdungeon -24, platapulse -8, Eisenhammer -6, black_sabbath -19, and 21 more...

@Aristoplato
Embrace the future. AI translations will only get better and better. People never appreciated human translations and took them for granted not understanding long hard work it is. AI gonna make make a lot of translators lose commissions, they probably need to find an actual job.

IchigoReader on GPT 5.0 is going to be insane. The current build can do an entire doujin under a few minutes. How on earth can anyone compete with that. Theoretically, a group can do all untranslated doujin ever released under a year.

Sooner or later, probably in a year's time people will literally download a doujin in E-hentai, upload it in a app and then read it under two minutes. Will the quality be good? In what way? Is it readable? It will be. Is it good enough to fap on? Probably. You're just gonna move on the newest one anyway. It will be like the COVID testing kits.

From what I have seen from IchigoReader 4o, people will have a hard time realizing that a lot of translators from the past aren't that good at all.

AI will only get better and better. In fact judging how bad "official" translations are. I would prefer them in the future if companies keep inserting wokeness in translations. Seriously why would you insert wokeness in a hentai translation?

What I want in for IchigoReader to become IchigoEditor where you can edit translations. Now that would be amazing.

Base +25, KorMeteor -6, black_sabbath -19, derp699 -6, Ingryo -6, H76NVa -6, psycomeGMK -7, Gintokiii -7, cipx -6, LingeringMemories -15, Gren92747 -7, EvilSanta -6, kailus13 -6, and 54 more...

yeah okay but the problem is that they're dogshit *right now*

i dont really give a shit if a platform wants to try to automatically translate on the fly. yeah, that's pretty much inevitable, you can get that built into your browser right now if you really wanted to

but we're still talking about a context-heavy language broken up into short sentences for visual flow, something that LLMs don't account for as-is. are there ways to get around that? yeah, but there's the thing with people uploading MTLs: they're not known for putting in any fucking effort. they're uploading garbage, en masse, without proofreading it, caring more about efficiency than actual output, or they otherwise don't even speak english and it's an enigma that they're even trying.

i dont need it to be poetry or ethically sourced or whatever, but if it's awkward, that shit takes you out of it immediately

if there was a way to easily tag and filter these out, nobody would even bother complaining about it anymore

Base +6, wtf616 +7, sonnytag123 +18, Britannia991 +7, Enduro +6, XunXunderedere +6, Black_Sand +6, Porokichin66 +9, bryci +11, winterpirates +6, LingeringMemories +15, Frank9969 +8, figzzz +6, and 21 more...

@Inspiretonic
If people accepted CGRascal translations as being actual translations, then this is more than dogshit. It's not perfect. Names are all over the place and bad word choices not to mention the editing that it doesn't translate handwritten moans but it's better than ever. At least these are translations, not rewrites.

Unless the doujin is complicated, then it is likely that the average doujin would be translatable by AI very soon. The thing is LLMs can account for sentences being broken up, it is the text retrieval of IchigoReader that fails in that scenario. I believe it translates bubble by bubble then hallucinates if it bubble is incomplete since the text is broken up.

Base +25, winterpirates +6, LingeringMemories -15, Soldier A -6, hobohobo -30, uanime5 +7, dankestdungeon -24, Aside +6, Darknessend -7, animadversion -6, EroYin -6, Jagyr -6, Meccaflare -8, and 9 more...

@shakuganaexa no, LLM can't account for sentences being broken up, because even humans can't account for it. And it's even worse in languages like Japanese where the subject is often omitted, because the context makes it obvious who or what and which gender the subject is.

When you have a japanese sentence that is basically "kissed the other person", you can have the best LLM possible, there is no way it will know if the sentence actually means "i kissed her", "he kissed me", or "she kissed her" or any other variation. You need context, you need to know the previous sentences, you need to see the image to see who spoke.

That's why MTLs invariably have stupid stuff like "I saw Mr. X" when talking about a girl, or stuff like "he needs to hurry" when it's actually a girl saying to herself "I need to hurry".

And with kanji having multiple reading, you often end up with those tools missing a proper name for a word, or translating the name differently on each page.

LLM can't account for that, and probably won't for at least a few more years. Even if you feed deepL multiple sentences at the same time, you'll get contradictions. They can yield good results if they are used as tools by people aware of their limitations and willing to crosscheck the results with the context and correct it accordingly. But that requires work.

Base +12, EroYin +6, Jagyr +6, LingeringMemories +15, Nivama +6, hydrati +6, Loli Poli +6, rvone +6, Porokichin66 +9, MurphysLaw +7, MehMeher +9, DJ_Puchuu +17, jzc136 +7, and 1 more...

As someone who cant understand japanese, a Machine translation is better than no translation, it would help if it had a (english) native speaker checking the actual translation to confirm it actually makes sense but hey, if real translators feelings get hurt and wont translate a work that has been previously translated, well, its a shame but it also seems like a personal problem, you dont see people complaining about machines replacing people at factories.

Base +6, MurphysLaw -7, MehMeher -9, Cocassu -7, TheTabbingMan +2, DJ_Puchuu -17, dankestdungeon -24, AwwGasm -10, TJsG4ng -7, lord-max -13