Hi, I'm the one who was uploading this (and my first post in the forum). At first, I thought it possible to upload anonymously, but that seems not the case. So please let me preface you with this: This is an MTL (if it's not obvious), and the "Machine" term is quite literal.
But enough ranting from me. The conclusion: Please do not expect me to upload the rest of the chapters. I don't have enough time, energy, or interest to develop and maintain to be something that I could do in long-term (I think other people on the internet did a better job at that anyway since this is entirely out of my field)
Lastly, If you think this violates a rule, please report it and take down the chapter if necessary.
(Sorry if I got the character names wrong, I do very minimum checks)
I'm a Software Engineer who just recently lived in Japan for about 5 months. But, despite this, I can't read or speak Japanese at all (only know some phrases that you may find commonly in anime/manga stuff) due to a lack the motivation/enthusiasm to learn the language. I've been looking to work on a side project for quite some time, so I decided to write a program that helps me motivate or at least correlates with both programming and learning the Japanese language. Thus, I wrote the program which automatically:
Note that this is far from my first (lazy) attempt. In fact, I commissioned some translation services before so I can compare the man-made translation/scanlation with the result of my tool (well, aside from the honest reason of instant access to enjoying the manga that no one has translated before, obviously lol). But the previous attempts were failures for obvious reasons: the text detection and translations are so garbage. It could barely detect anything correctly or translate anything that can be comprehended.
However, ChatGPT / LLM has come into my mind occasionally since a few months ago to continue this side project since they're getting better and better every day. Thinking about powering my tool with it to help with the translations and OCR stuff.
The result? well, this chapter is the result which I think is bearable to read. Why this manga? well, it was on top of the list of manga I'm interested in that got abandoned. About 95% of the work was my tool. The rest were small manual adjustments which consisted of "Feeding" the story context to the ChatGPT, Purchasing the raw and making it accessible to the scrapper, small corrections on misdetection results, applying formatting, and adjusting SFX / Bubble positions all inside the JSON files, Oh and lazily fixing some artifacts. I don't even have Photoshop installed on my workstation.
But despite all of these, I wasn't proud of the quality, and I didn't gain any motivation boost with the programming and language learning that I was looking for.
Note: I don't plan to expose this disastrous source code anytime in the future, so please don't ask.
- Scrap the images from the raw sites (and upscale them if necessary, because the raw of this image is goddamn small)
- Detect the text and dump it into JSON files (so the transcripts can be modified by hand. e.g. apply formatting bold, italic, etc. or do simple transformations when text overlaps each other)
- Clean the bubble text from the images
- Laid the translated text into the cleaned bubble / overlay text for the SFX
Note that this is far from my first (lazy) attempt. In fact, I commissioned some translation services before so I can compare the man-made translation/scanlation with the result of my tool (well, aside from the honest reason of instant access to enjoying the manga that no one has translated before, obviously lol). But the previous attempts were failures for obvious reasons: the text detection and translations are so garbage. It could barely detect anything correctly or translate anything that can be comprehended.
However, ChatGPT / LLM has come into my mind occasionally since a few months ago to continue this side project since they're getting better and better every day. Thinking about powering my tool with it to help with the translations and OCR stuff.
The result? well, this chapter is the result which I think is bearable to read. Why this manga? well, it was on top of the list of manga I'm interested in that got abandoned. About 95% of the work was my tool. The rest were small manual adjustments which consisted of "Feeding" the story context to the ChatGPT, Purchasing the raw and making it accessible to the scrapper, small corrections on misdetection results, applying formatting, and adjusting SFX / Bubble positions all inside the JSON files, Oh and lazily fixing some artifacts. I don't even have Photoshop installed on my workstation.
But despite all of these, I wasn't proud of the quality, and I didn't gain any motivation boost with the programming and language learning that I was looking for.
Note: I don't plan to expose this disastrous source code anytime in the future, so please don't ask.
But enough ranting from me. The conclusion: Please do not expect me to upload the rest of the chapters. I don't have enough time, energy, or interest to develop and maintain to be something that I could do in long-term (I think other people on the internet did a better job at that anyway since this is entirely out of my field)
Lastly, If you think this violates a rule, please report it and take down the chapter if necessary.
(Sorry if I got the character names wrong, I do very minimum checks)
