For those unfamiliar, 99% Invisible + Hidden Levels did an episode about the backstory of development of Segagaga that is really worth a listen. I really enjoyed it... and without having listed to it months ago wouldn't have even known what this headline was about!
This is the first time Dreamcast owners can play it in a more understandable way (and I don't understand why some users are boycotting the translation due to the AI used in the translation work, but that's beside the point). Playing SGGG, we can all see how brave Sega was at the time. I'm still shocked that Sony had such good marketing around the PS2 launch, so users could trust them to wait - and NOT buy - a Dreamcast. The Dreamcast had already great games (Jet Set Radio! MSR! Shenmue! RECV! and many more!), great online features, and its architecture was so well thought out.
> and I don't understand why some users are boycotting the translation due to the AI used in the translation work, but that's beside the point
Segagaga has a ton of obscure, referential, meta humor that isn't easily translated to English. The "cleaned up machine translation" approach means that a lot of this is lost. Looking at some screenshots of the game, the script seems stiff and overly formal, much like how direct machine translation of Japanese text reads. https://bsky.app/profile/did:plc:2jjromh55tf7pp7s4hsvurf4/po... Obviously it's better than nothing, but people are pissed off because the "edited machine translation" workflow leads to poor results, not because of some reflexive anti-AI bias or whatever.
Based on what I read in the article, MT was used as an experiment and proof of concept, and dedicated translation was done for the full release. Was that not the case?
Was the workflow MT -> human translation?
If so thats honestly such a lazy followthru given the technical hurdles overcome wrt the font tech
From the Github repo (https://github.com/ExxistanceDC/Segagaga-English-Translation), the translation went through a process called MTPE (Machine translation, post-editing). This works just like it sounds. The initial translation is done with machine translation, then human translators review and edit the resulting translation to try to correct any mistakes.
> What I call the “playtesting translation” — a base translation that allowed the artists and playtesters to get started early and understand what they were working on — was developed using a combination of DeepL and ChatGPT 4o/4.5. That translation then went through a substantial, months-long human translator review. I don't think that the end product feels “machine-translated,” but that’s ultimately for you, the player, to judge.
> MTPE (Machine translation, post-editing). This works just like it sounds. The initial translation is done with machine translation, then human translators review and edit the resulting translation to try to correct any mistakes.
And the consensus among professional translators is that MTPE only saves time if you're willing to accept a half-assed result. For them to edit MT up to the standard of manual translation takes just as much expertise and effort as translating it manually in the first place.
> And the consensus among professional translators is that MTPE only saves time if you're willing to accept a half-assed result.
I have no particular interest in translation, but clearly when the person saying X is bad depends financially on you not buying X, you must take their word with a grain of salt.
To be fair what they're saying holds up intuitively. To recognize every possible case where the MT may have deviated from what a human translator would do, the human editor has to translate it themselves anyway for comparison. Best case is they save a bit of typing if the MT actually nailed it first try.
In practice what the industry is hoping for is that under sufficient time and pay crunch, the translators will have to let their standards slip and rubber-stamp "close enough" machine translations rather then going the extra mile.
> 12.08% [of translators] say MTPE produces high-quality output.
> A significant portion (around 50%) of respondents do not offer discounts for MTPE work, arguing that post-editing can take as much time as traditional translation.
> Among those who do offer discounts, the most common range is between 10-30%.
As a side bar, I found it funny how in a Stanford lecture teaching various routes to training llms on the Machine Translation benchmark, the sample used was a French to English translation of 'the teddy bear is blue' or something similar.
After the lecture I reviewed the current production grade google translate...it butchered the translation
i don't think the translators had any first hand experience with akiba culture of that time period, nor did they know how to speak japanese to actually research it (and most of those ezweb.ne.jp / imode sites are long gone)
Is it possible for AIs in the future catch up those gags and meta humor in the language? Could the translation team have done a better job prompting the AI with the current technology?
Afaik japanese media (manga) abuse alot the use of puns on its language due to the amount of homonyms on it.
Avante-garde will always be lost on an inferential technology because to 'get it' it needs to be trained on sufficient volumes.
67 67 67 67 67 67 67 67 67 67
Edit: I may not exactly be spitting pearls, but passive downvoting for a genuine take on the matter based on professional experience in AI is a canary gasping for oxygen.
As someone who owned a dreamcast and stayed in line all night for a PS2, I can say that the DVD player functionality was HUGE. It was many, many people's first DVD player, and priced so that the game console was nearly 'free' on top of it.
The Dreamcast still got played post-PS2 purchase, but not much - fighting games, mostly. PS2's catalog was very, very strong and the combo just dominated.
It was much easier to pirate games on the Dreamcast - the copy protection was broken to the degree that you could burn a game to a CD-R and have it Just Work without modifying the console in any way. It being both a total piracy free-for-all and also a catastrophic commercial failure doesn't seem to fit what you're saying.
Not to say that easy piracy is necessarily a death sentence for a console, the DS succeeded in spite of ubiquitous and cheap flashcarts, but the Dreamcast shows it's not necessarily a path to success either. There are just more pertinent reasons for a system to sink or swim.
I think that it was a game player and a DVD player had more to do with the success of the thing. Oh and it plays psx games too.
I owned both. The graphics/games were of similar quality. Having a larger game storage gave the ps2 a decent advantage. The dreamcast seemed more interesting. But the PS2 had a better customer feature set.
Lol what? You got it the other way around. Also out of the dozen or so friends in my friend group who had a PS2, none of us had a modded PS2 or pirated games.
The PS2 was popular on its own and it wasn't related to piracy.
What are you talking about? The Dreamcast didn't even have adequate copy protection. I switched my entire collection to backups at some point to preserve the integrity of the original discs.
On one hand, I'm happy this got translated and love the dreamcast. On the other, this game is so weird I doubt the translation holds a candle to the original
It's a shame to see all the pearl-clutching about the use of AI. I've been playing ROMs and fantrans since the early aughts, and we were content back then when some of our desired games just got menu translations. I've watched a ton of projects spring up around the games people wanted to play, only to stall or die out because the majority of the project people were programmers, and the one translator was a college student enrolled in Japanese 102 who flaked out.
Human translation should obviously be the end-goal, particularly in a text-heavy game from the 32-bit era...but that shouldn't undercut the technical achievement of this hack, even if the MTL text becomes little more than placeholder. Put it all up online, make it easy for someone to pick up the script and translate it independently in chunks, and then insert it back in later. That's how a number of fan-trans were done in the past, if memory serves.
Also AI for translation is one of the best and oldest uses of AI, to the point where if this translation didn't exist, you could probably get by simply pointing your phone at the text and getting real time translations. An open forum for communicating back all the places people go "huh, that doesn't seem right" and fixing it is even better.
Amazing technical achievement, and also a very unfortunate situation with the reliance on AI translation. I'm not privy to the details of the actual translation process (whether AI translation was really just a technical placeholder or actually the draft for the real translation), but the criticism online has been scathing.
Yes the real achievement here I think is the creation and public sharing of the translation tools. The AI translation here should be considered effectively a proof of concept of the tools.
Given the particular nature of this game, so reliant on inside jokes requiring a knowledge of SEGA history, it's likely an AI translation could miss a lot, and I think the community will eagerly await further real translations done by professional translators leveraging these tools.
I hope so... but in the fan translation world the first translation is almost always the only translation, and the only one most people will experience even in the exceedingly rare case of an alternate translation coming into existence (as only the first one ever makes the news).
The article says "Something important that should be clarified before we move on is that machine translation (specifically a combination of DeepL and ChatGPT 4o/4.5) was used during the initial hacking phase for testing purposes. Once it was time to translate the game proper, multiple human translators then took over."
Fan translations have used Babel Fish or similar during the development process for decades. If the final script isn't AI translated I don't see any issue with that.
This statement still, unfortunately, doesn't clarify whether the final translation was MTPE (machine translation post-editing) or translated from scratch by humans. I'm not a professional translator, but according to those of them I've seen on Bsky and the likes, MTPE greatly hurts both the translation quality and, ironically, the translation speed. It also makes it easier to accidentally recruit translators who don't have a decent enough grasp of the source language.
Screenshots of the translated game do give an impression of edited - or even, at times, unedited - machine translation. What with overly direct word-by-word translations, as well as reasonably obvious references (which the game is chock full of) getting mistranslated as something else entirely. Although those screenshots, of course, are not necessarily representative of the whole script, which was a collaborative effort by many translators.
> this really was the white whale translation of the Dreamcast library, there was a lot riding on this project
I don't understand... what is riding on this?
I have never heard of this game, but based on this blog post I can see why it's a "white whale". It's precisely why I don't think this was a good use of AI at all.
For those unfamiliar, 99% Invisible + Hidden Levels did an episode about the backstory of development of Segagaga that is really worth a listen. I really enjoyed it... and without having listed to it months ago wouldn't have even known what this headline was about!
https://99percentinvisible.org/episode/hl-06-segagaga/
This is the first time Dreamcast owners can play it in a more understandable way (and I don't understand why some users are boycotting the translation due to the AI used in the translation work, but that's beside the point). Playing SGGG, we can all see how brave Sega was at the time. I'm still shocked that Sony had such good marketing around the PS2 launch, so users could trust them to wait - and NOT buy - a Dreamcast. The Dreamcast had already great games (Jet Set Radio! MSR! Shenmue! RECV! and many more!), great online features, and its architecture was so well thought out.
> and I don't understand why some users are boycotting the translation due to the AI used in the translation work, but that's beside the point
Segagaga has a ton of obscure, referential, meta humor that isn't easily translated to English. The "cleaned up machine translation" approach means that a lot of this is lost. Looking at some screenshots of the game, the script seems stiff and overly formal, much like how direct machine translation of Japanese text reads. https://bsky.app/profile/did:plc:2jjromh55tf7pp7s4hsvurf4/po... Obviously it's better than nothing, but people are pissed off because the "edited machine translation" workflow leads to poor results, not because of some reflexive anti-AI bias or whatever.
Based on what I read in the article, MT was used as an experiment and proof of concept, and dedicated translation was done for the full release. Was that not the case?
Was the workflow MT -> human translation?
If so thats honestly such a lazy followthru given the technical hurdles overcome wrt the font tech
From the Github repo (https://github.com/ExxistanceDC/Segagaga-English-Translation), the translation went through a process called MTPE (Machine translation, post-editing). This works just like it sounds. The initial translation is done with machine translation, then human translators review and edit the resulting translation to try to correct any mistakes.
> What I call the “playtesting translation” — a base translation that allowed the artists and playtesters to get started early and understand what they were working on — was developed using a combination of DeepL and ChatGPT 4o/4.5. That translation then went through a substantial, months-long human translator review. I don't think that the end product feels “machine-translated,” but that’s ultimately for you, the player, to judge.
> MTPE (Machine translation, post-editing). This works just like it sounds. The initial translation is done with machine translation, then human translators review and edit the resulting translation to try to correct any mistakes.
And the consensus among professional translators is that MTPE only saves time if you're willing to accept a half-assed result. For them to edit MT up to the standard of manual translation takes just as much expertise and effort as translating it manually in the first place.
> And the consensus among professional translators is that MTPE only saves time if you're willing to accept a half-assed result.
I have no particular interest in translation, but clearly when the person saying X is bad depends financially on you not buying X, you must take their word with a grain of salt.
To be fair what they're saying holds up intuitively. To recognize every possible case where the MT may have deviated from what a human translator would do, the human editor has to translate it themselves anyway for comparison. Best case is they save a bit of typing if the MT actually nailed it first try.
In practice what the industry is hoping for is that under sufficient time and pay crunch, the translators will have to let their standards slip and rubber-stamp "close enough" machine translations rather then going the extra mile.
> consensus among professional translators
[citation needed]
https://blog.gts-translation.com/2025/04/07/the-state-of-mac...
> 12.08% [of translators] say MTPE produces high-quality output.
> A significant portion (around 50%) of respondents do not offer discounts for MTPE work, arguing that post-editing can take as much time as traditional translation.
> Among those who do offer discounts, the most common range is between 10-30%.
Oh boo hoo.
Just boooooo.
Disappointed. Gif
As a side bar, I found it funny how in a Stanford lecture teaching various routes to training llms on the Machine Translation benchmark, the sample used was a French to English translation of 'the teddy bear is blue' or something similar.
After the lecture I reviewed the current production grade google translate...it butchered the translation
I wouldnt trust machine translation.
i don't think the translators had any first hand experience with akiba culture of that time period, nor did they know how to speak japanese to actually research it (and most of those ezweb.ne.jp / imode sites are long gone)
Is it possible for AIs in the future catch up those gags and meta humor in the language? Could the translation team have done a better job prompting the AI with the current technology?
Afaik japanese media (manga) abuse alot the use of puns on its language due to the amount of homonyms on it.
Avante-garde will always be lost on an inferential technology because to 'get it' it needs to be trained on sufficient volumes.
67 67 67 67 67 67 67 67 67 67
Edit: I may not exactly be spitting pearls, but passive downvoting for a genuine take on the matter based on professional experience in AI is a canary gasping for oxygen.
As someone who owned a dreamcast and stayed in line all night for a PS2, I can say that the DVD player functionality was HUGE. It was many, many people's first DVD player, and priced so that the game console was nearly 'free' on top of it.
The Dreamcast still got played post-PS2 purchase, but not much - fighting games, mostly. PS2's catalog was very, very strong and the combo just dominated.
The reason the PS2 was successful it was because it was very easy to unblock it and use pirated games.
The Dreamcast wasn't as easy as I can remember.
It was much easier to pirate games on the Dreamcast - the copy protection was broken to the degree that you could burn a game to a CD-R and have it Just Work without modifying the console in any way. It being both a total piracy free-for-all and also a catastrophic commercial failure doesn't seem to fit what you're saying.
Not to say that easy piracy is necessarily a death sentence for a console, the DS succeeded in spite of ubiquitous and cheap flashcarts, but the Dreamcast shows it's not necessarily a path to success either. There are just more pertinent reasons for a system to sink or swim.
I think that it was a game player and a DVD player had more to do with the success of the thing. Oh and it plays psx games too.
I owned both. The graphics/games were of similar quality. Having a larger game storage gave the ps2 a decent advantage. The dreamcast seemed more interesting. But the PS2 had a better customer feature set.
In models built prior October 2000 - it was very easy, just boot up Utopia loader and then you were able to run any game from CD.
Lol what? You got it the other way around. Also out of the dozen or so friends in my friend group who had a PS2, none of us had a modded PS2 or pirated games.
The PS2 was popular on its own and it wasn't related to piracy.
What are you talking about? The Dreamcast didn't even have adequate copy protection. I switched my entire collection to backups at some point to preserve the integrity of the original discs.
> boycotting the translation due to the AI used in the translation work
Non-AI translators had literal decades to step up and do this and didn't so honestly who cares what they think about it.
On one hand, I'm happy this got translated and love the dreamcast. On the other, this game is so weird I doubt the translation holds a candle to the original
So excited for this - have been dying to play it for at least a decade.
Now if we can just get a fan translation of London Seirei Tanteidan (PS1 RPG set in victorian England).
It's a shame to see all the pearl-clutching about the use of AI. I've been playing ROMs and fantrans since the early aughts, and we were content back then when some of our desired games just got menu translations. I've watched a ton of projects spring up around the games people wanted to play, only to stall or die out because the majority of the project people were programmers, and the one translator was a college student enrolled in Japanese 102 who flaked out.
Human translation should obviously be the end-goal, particularly in a text-heavy game from the 32-bit era...but that shouldn't undercut the technical achievement of this hack, even if the MTL text becomes little more than placeholder. Put it all up online, make it easy for someone to pick up the script and translate it independently in chunks, and then insert it back in later. That's how a number of fan-trans were done in the past, if memory serves.
Also AI for translation is one of the best and oldest uses of AI, to the point where if this translation didn't exist, you could probably get by simply pointing your phone at the text and getting real time translations. An open forum for communicating back all the places people go "huh, that doesn't seem right" and fixing it is even better.
Amazing technical achievement, and also a very unfortunate situation with the reliance on AI translation. I'm not privy to the details of the actual translation process (whether AI translation was really just a technical placeholder or actually the draft for the real translation), but the criticism online has been scathing.
Yes the real achievement here I think is the creation and public sharing of the translation tools. The AI translation here should be considered effectively a proof of concept of the tools.
Given the particular nature of this game, so reliant on inside jokes requiring a knowledge of SEGA history, it's likely an AI translation could miss a lot, and I think the community will eagerly await further real translations done by professional translators leveraging these tools.
I hope so... but in the fan translation world the first translation is almost always the only translation, and the only one most people will experience even in the exceedingly rare case of an alternate translation coming into existence (as only the first one ever makes the news).
The article says "Something important that should be clarified before we move on is that machine translation (specifically a combination of DeepL and ChatGPT 4o/4.5) was used during the initial hacking phase for testing purposes. Once it was time to translate the game proper, multiple human translators then took over."
Fan translations have used Babel Fish or similar during the development process for decades. If the final script isn't AI translated I don't see any issue with that.
This statement still, unfortunately, doesn't clarify whether the final translation was MTPE (machine translation post-editing) or translated from scratch by humans. I'm not a professional translator, but according to those of them I've seen on Bsky and the likes, MTPE greatly hurts both the translation quality and, ironically, the translation speed. It also makes it easier to accidentally recruit translators who don't have a decent enough grasp of the source language.
Screenshots of the translated game do give an impression of edited - or even, at times, unedited - machine translation. What with overly direct word-by-word translations, as well as reasonably obvious references (which the game is chock full of) getting mistranslated as something else entirely. Although those screenshots, of course, are not necessarily representative of the whole script, which was a collaborative effort by many translators.
This is absolutely amazing and I can't wait to deep dive on this.
> this really was the white whale translation of the Dreamcast library, there was a lot riding on this project
I don't understand... what is riding on this?
I have never heard of this game, but based on this blog post I can see why it's a "white whale". It's precisely why I don't think this was a good use of AI at all.