Scanlation Thread

Any translator interested in working on some h? specifically /ss/

Do any of you guys use programs like tinypng to decrease file size?

Yes. Bandwidth and storage is cheap and plentiful.
Why would sacrifice quality for thirdworlders using shit tech?

Third worlders have better internet, dummy.

This is already dumb limit and will be even dumber when modern image formats get adopted.

I don't think phone users can really tell. Something about the way phones display images.

Every time I've done that, it only shaved 1-2% off of the file size. Save for Web with 256 colors seems to work just fine.

pnggauntlet for lossless compression via pngout, optipng and whatever else it is using. My photoshop export settings usually turn 1400px sites into ~800kB files and pnggauntlet shaves 2-400kB off. No lossy compression, the digital raws are bad enough as they are.

unless they zoom in phones do the same thing as computers, they simply downsample the image to the output resolution. Also not all phones are ~1080p, higher end models have been going up to ~1440p since like half a decade

Here's what I got for savings on a chapter I was working on. I have no idea how my images were so ineffecient.

It went from 22.8 to 14.2 MB.

Attached: savings.png (180x524, 7K)

Are you sure you're using lossless?

It was tinypng's default settings.

I'd check in photoshop to be sure, default settings will fuck you.

The current standard is already 1440p which means phones and tablets have a height of 2560px so yes, the bare minimum should be 1600px but no the standard.
It should definitely be higher than 1600px.
People complaining about bandwidth or space are just faggots who don't know how to use PNG properly.
And if someone starts complaining about downscaling, in case it's even higher, tell him to learn how to program the viewer properly.

Setting a limit is retarded, higher is always better. Especially in prospects toward the future.

At some point, with larger resolutions, you're seeing the page in a manner that's similar to viewing the physical page with a magnifying glass.

Why? My display is 2880x1800 so I can handle much higher than 1600px.

>When you upload a PNG (Portable Network Graphics) file, similar colors in your image are combined. This technique is called “quantization”. By reducing the number of colors, 24-bit PNG files can be converted to much smaller 8-bit indexed color images. All unnecessary metadata is stripped too. The result better PNG files with 100% support for transparency. Have your cake and eat it too!

He should check how the colors are.

Back when I were still scanlating, FileOptimizer was THE best program to optimize images. It used shitload of the latest third-party utilities and provided the best results. I was autistic enough to spend a couple of days researching and this bad boy was unbeatable.

It has nothing to do with the display, it's about how the art was originally drawn by the artist.

I saw that, but the website was so shady looking I decided not to.

nikkhokkho.sourceforge.io/static.php?page=FileOptimizer

For most I've seen you see fidelity improvements well beyond 1600px.

>most displays these days have at least 1440p resolution and we move towards 4K
>file sizes for manga basically don't matter with todays storage spaces and internet speeds
>bigger resolutions provide more detail and can easily be scaled down

somehow Yea Forums thinks quality with virtually no downsides is bad

because manga artists down't draw on a shitty 1600px grid, as can be seen in good prints or digital releases (like my senpai is so annoying)

github.com/fhanau/Efficient-Compression-Tool
Is what I'm using. Has multithreading.

just do it yourself my dude
throw everything into google translate, fuck with the wording somewhat so it's vaguely coherent even if it's wrong, and typeset
it's porn, barely anyone gives a shit

I know. Which was my point.
The source is higher resolution and displays are as well so why would it be like using a magnifying glass.

Google translate doesn't work for asian languages, all you get is literal nonsensical gibberish
>it's porn, barely anyone gives a shit
I give a shit...

Funny secret. I'm a "translator", but I don't know Japanese. I use my phone and its google translate function to output the romaji. I also make good use of its kanji recognition capabilities. I then use online dictionaries and common sense to figure out what the sentence is.

No one suspects a thing and even people who know Japanese think my "translations" are great.

if you're putting in that much effort you might as well learn the language so it isn't dead time

Yes.
I hate that fag who uploads 90000 pixels of Anjou-san. It makes my laptop lag all the time.

You're just lying to seem like le epic troll. You might do rewrites by making up the plot yourself, but you're not going through all that trouble considering how bad google translate is.

Google, romajidesu, and other tools go a long way. But I do have to google certain phrases and slang I can't find. I can usually translate a short chapter in a day or two this way.

What app does the ocr? What's your process?
I'd do this but I don't know how to type the moonrunes into g translate.

Yeah.. EOP won't be able to spot a single mistranslation as long as you have perfect English grammar.
Worthless EOP never bitch about accuracy since they're too incompetent to do that.

>"laptop" can't handle ~60MB of manga
might be time to switch that calculator for something from the current milennium

>it's porn, barely anyone gives a shit
Everyone here only works on shitty harem and isekai and shojo garbage. I'm not kidding when I say that porn manga is usually better written.

You don't really type anything. The google translate app on your phone makes use of your camera. It circles the text and will output it in romaji. If you scroll over kanji it will tell you want it means if it's in it's dictionary.

From there I break the sentence apart and figure out each word. Romajidesu is very good for this, since you compare your manga page to the variations of how the word is spelled.

If you can't type Japanese characters, just find a site that allows you to copy and paste them when needed.

You can also use your phone to try to draw the kanji in question with your finger on your screen and it will give suggestions on what that might be.

>average isekai translator

So you take picture of computer screen with phone?

Anything less is unacceptable.

Attached: 005.jpg (4228x6071, 3.19M)

Yes. It doesn't work though, the guy is just yanking your chain.

Porn is harder to translate because it's full of hai reberu vocabulary for adults.

I buy the physical magazine, since it's easier to scan from. You have to open the google translate app.

It works and I've been doing it for months. It just takes time and looking up some things online.

I suppose that explains why the translations for Sekkaku Cheat are awful.

>If you have access to scans of suitable quality all releases should be like anjou san with 3000x4200 and ~8MB per image.
I just checked, but why the fuck do they save it in jpg? They can easily go down by half of the size by using good png settings.

Attached: M2.png (3005x4227, 3.98M)

Are you retarded? Do you even own or scan physical manga? A 600 dpi scan at full resolution looks like the equivalent of putting a page a few cm from your face. You see a lot of intended details, such as the fibers of the page, stray ink marks, etc. Analog vs Digital is a tricky subject because obviously the in real world, each page has "infinite" resolution. You can look at it from far away or with a magnifying glass, and the detail will scale accordingly. With digital resolutions, you capture how the page looks at a specific distance and any zooming in or out will create distortions.

didn't notice that, you're right. Don't know why they do that either

if you have shitty digital 1200px releases is it acceptable to upscale the image by 2 just to get some space for typesetting and making the font look less pixelated? Or is the difference between clean text and pixelated to shit art more of an eyesore?

>A 600 dpi scan at full resolution looks like the equivalent of putting a page a few cm from your face.
Only on your shitty monitor with low pixel density.
If you had a trillion pixel scan and a trillion pixel resolution tablet it'd just look like a super sharp scan/paper.

good job blurring your images.

people who upscale should be shot

No. Even if they're shitty 1200px, at least (in most cases) they're still sharp. If you upscale them you destroy even that little bit of "quality" that's left over.

You mean you don't use waifu2x?

but upscaling by two changes absolutely nothing. Instead of one pixel of a specific color value you just have four

>Only on your shitty monitor with low pixel density.
That has nothing to do with the details captured by a scan, you fucking retard.

Read this manga and see how awfully blurry these shitty upscaled images are.
mangadex.org/title/22615/onii-chan-is-done-for

>upscaling with nearest neighbor
oh-shit-nigger.png

Resizing is ok, but you should always make the original size available. Not everyone is a fucking indian or american with data caps.
On my hand, in the future i'll release a oneshot from a famous series at 4500h, 10mb per page of uncompressed png.
Anything under that is literally for dumb nigger phone posters.

Harta, AX, Kirara, anything that doesn't have a digital version really

Even if your super scan captures every grain of paper pulp if it's displayed on a roughly equivalently sized display with high pixel density you won't notice these details.
If I scan a 9 inch manga page at XXX pixels and display on 10 inch tablet of XXX pixel resolution it's not going to look weird.

All Kirara magazines are digital now. Two of the four are just on a delay.

All Kirara are digital now.

Americans don't have data caps, generally speaking. You're thinking of yuropoors and strayans.

>yuropoors
Bongs maybe, they're not really euros.

there's no such thing as data caps in europe, for most home internet connection
for mobile, you can get a 50GB for 50-10€ easily everywhere

yes, looks like shit. But if i upscale by an even factor without applying any filters or anti aliasing it will not blur. Worst case is jaggedness when people zoom in

i didn't know that's what it was called. Willing to tell me what's wrong with it?

Attached: Untitled-1.png (8x4, 115)

is there a digital version of Kindai Mahjong magz?

Keep seething, yuropoors. Notice how you randomly started sperging about America. Why don't you go to your own yuropoor imageboard? Oh right, you don't have one.

Americans have caos sometimes. Bongs have caps. Rest of yurop doesn't have caps and is 5 times faster than the fastest murrican internet.

You have five gigabit internet?

Does anyone know if there's a program/script to rip from Comic-Trail ( comic-trail.jp/ )? The pages aren't encrypted, but they're served as some separate blob-files, making it a pain to save all of them by hand. They have some series I'd like to try my hand at at some point, so I figured I'd ask if something existed before making something myself that might already exist.

It looks like shit is what's wrong with it. You're right in saying that it doesn't blur, but it makes everything look pixelated, even without zooming in. It's really no better than something like bilinear.

And then you slap the text on it in double the DPI and give everyone a headache.

That's easy. People have been writing a ripper some threads ago that uses the same method.

Image: comic-trail.jp/pt/atari_00001_V8fjmaiE/data/0003.jpg
Scramlbing: comic-trail.jp/pt/atari_00001_V8fjmaiE/data/0003.ptimg.json

Etc.
I didn't save the script but maybe some user who didn't miss the expiring pastebin did and you're in luck.

Thanks for the tip. I got decent compression with pnggaunlet.

bilinear blurs line art. I don't see how nearest neighbour upscaling by 2 would make it look pixelated as long as you display it filling the same screen as the original image. It only becomes a problem if you zoom in (granted, some people will do that, and for people with >1080p vertical resolution it will fill the screen looking like shit).

IMO you should always export as high as possible. It should be up to the online reader software to automatically scale down to accommodate different devices (which I have specifically built mine to do).

>as long as you display it filling the same screen as the original image
Well, then there's no point in upscaling, is there?

thanks for pointing out my retardation. Yes, sure, makes sense. I just thought i could make things like pic related look better but in order to see a difference you'd have to zoom in at which point everything else would look even worse. Oh well, at least now i know.

Attached: Capture.png (357x308, 17K)

Uh, sure, any time.

>It should be up to the online reader software to automatically scale
Enjoy your moire patterns

I'm not retarded. It uses Mitchell or Lanczos

Ten.

Enjoy your moire patterns.

Once again, never had a problem

This was made by someone for another site but also works for this, I put it up for 10min pastebin.com/bsgujdtR
Without changing the save location mine seem to disappear into the void, might want to change that.

Just make something yourself, that’s one of just a handful of readers that’s used all the time and it’s not very hard.
I just woder if there’s a way to get the images pre-scrambled, that tends to have a bit higher quality. But if they don’t offer an app or have some giant security holes, unlikely.

The heck? I don't even get a scrambled image like that, it's just every page divided in three strokes. Maybe because I've been experimenting with the free chapters they put on trial?

Thanks, got it. Should be able to reverse-engineer it to do what I want, even if it doesn't work.

I had made a start, but figured I'd check if someone else did it, first.

>I don't even get a scrambled image like that, it's just every page divided in three strokes
That's just how your browser keeps it after it's been descrambled.

You're in luck, I'm the guy who wrote that specific ripper. Just checked and it works with no changes needed for this particular site.

pastebin.com/z2yDEhQE

Just plug in the base url as an argument to the python:
script.py comic-trail.jp/pt/atari_00001_V8fjmaiE

People on smaller devices can use zoom and scroll if the manga is at high resolution. Mobile manga readers are designed for them to easily be able to read the pages.

As for people who are on 1080p computer screens and below, they can use fit to width to read at a size where everything is at a good size.

It recently occured to me how bad preservation of raw scans is. Some things are nowhere to be found even though they were uploaded on Share, PD, DDL sites, even nyaa. Dead links, dead torrents and incomplete files all over. Too bad there's no central site like sadpanda.

That's an older version, but it's close enough to the final do mainly the same things. Ideally, it should be making a folder in the directory that the script was run in, but in the older version, it doesn't account for non-UTF-8 encoding, so it'll try and make folder names with Shift-JIS encoding, which doesn't work too well. That may be the issue there, but I couldn't pinpoint the problem without more information.

Picture of the output.

Attached: Output.jpg (1032x923, 358K)

My main problem with ripping is that I have no idea where the fuck stuff is, what sites offer apps and so on. Or, when they offer apps, which apps offer better, the same or worse quality. Figuring all that stuff out takes more time than actually writing a ripper.

I don't believe you.

Probably invalid NTFS characters or some Windows bullshit like that.

Isn't that obvious? Reversing the app and figuring out what it does takes more time than writing it all up afterward.
The only way to know which is better is to simply check them all out.

It's sad. I don't get why goddess disallowed raws for awhile.

It's easy to reproduce yourself. Get libvips, downsize with lanczos. For faster downsizes, I use golang's mitchell resizing in my api. Still haven't had any issues. This is from 8k and other various height sources down to 2048, 1600, 1280 etc.

No, I mean figuring out if the app is worth reversing in the first place is a fucking pain and already at least 80% of the work.

My version was from before it created a folder, I assumed it would dump everything to the folder the script was located in. But either that's not the default functionality of save() or something was fucked up.
I try to archive some, although I mostly download things when I find a convenient batch instead of getting everything one by one.

Attached: 1529237693910.jpg (261x27, 7K)

I'm still waiting for them to allow raws again so I can dump many many terabytes on them.

I thought they allow it now?

Post proof. I don't believe you.

well give me a source image then

citation needed. I'm still going off of state-of-the-goddess.txt and last time I asked one of the mods

Am I missing something or does mangadex not have a contact staff button?

Which reminds me, a few months ago someone was complaining about them having a lot of useless duplicate chapters that take space, but now they use over 7 TB again.

They have a discord server, a public forum and private messages, what else do you need?

Only good site I know that's not a fucking blog is moeshare but it's a chink forum with baidu links. jcafe is still down and barely alive anyway. animebytes is kinda meh.

Raws can be deduplicated either by their epub identifiers or file hashes

Never mind you're right.
Anyway I can get those raws? I'll mail you a hdd even.

You're going to need more than one. I haven't finished deduplication yet, so right now its around 30TB, and growing. Busy building a search engine for it all

Why is it so much larger than goddess?
Search engine?

Where do you even get all of that?

Attached: 1532680991205.png (491x391, 37K)

> Why is it so much larger than goddess?
Because the goddess won't accept my offerings

> Search engine?
Yes, because the data is far too large to search manually. Being able to sort and search using metadata is very convenient, and useful for doing statistics.

Lots of work

It doesn't help that Japanese publishers go crazy and try to shit the pool by uploading shit manga scans fast so that it "encourages" people to buy the physical thing.

Thanks user. I hadn't worked with Python in years, so it took a while to get it to work. Turns out I needed Python >3.6, whereas I only had 2.7. Installed 3.7, and a couple of pip-installs later, and it worked like a charm.

What are your tips for newbies?

Mine would be:

1) Try out all group roles and figure out which ones you can do.

2) Try to contact a group or start a group so you can make contacts and have other people fill the roles you can't fill yourself.

3) Get proofreaders.

4) Try to avoid resizing your scans to preserve quality.

5) Save scans as .png files, since .png is lossless.

6) Use lossless compression apps, like pnggauntlet to further reduce .png file size.

Center-justify your typesetting.

Wouldn't that not work in cases where you have awkwardly shaped dialogue bubbles.

Do alot of people not do this?

>6) Use lossless compression apps, like pnggauntlet to further reduce .png file size.
No. Use Save-for-Web.

Where would a translator go to find raw light novels and manga?

It's not common, fortunately, since it's something that no typesetter worth a damn should fuck up, but it seems like such a simple thing that there's no excuse for it ever happening. I don't care if this is your first attempt and you've never typeset a day in your life beforehand, at the very least you should be able to tell when the way your text is justified looks like shit.

Alot of sites that I know of are missing chapters unless it's some super popular shonen magazin or something. I think alot of people just crack the manga apps and ebooks nowadays.

1. find good manga
2. translate
3. typeset
4. make a credit page "please continue the scanlation. thanks. bye."
5. upload

I know people that use save for web and they still get further reduction by using the app as well.

This completely ignores cleaning, redrawing, and proofreading.

>find good manga
Most stuff I want to translate is already taken by other groups.

out of curiosity i just tried out
>convert *.jpg out.png
and it changed the chapter from 85.2MB to 56.3MB
running optipng on that (lossless compressor) is taking a minute but it's doing about another 5% compression on average, so that'd be 53.5MB

in their defense JPG does usually do smaller filesizes in most other usecases, but they should've double checked, this is the type of case where PNG shines

also in their defense, they did complain for a few chapters about not having a good cleaner, which is why releases take so long to come out for Anjou.

How many pages was the chapter?

Just snipe it, they don't have a license, they dont own it.

11 pages, 12 if you split the spread.

My tip is to do everything yourself except maybe finding raws desu

optipng just finished, it did better than that actually. 52.5MB.
seven of the eleven files are still not postable on Yea Forums, though, so that still sucks. I'll still have to downsize 15-1600px next time i dump.

11-12 pages
56.3MB

How big are your scans? My short chapter at 2050 is only 21MB.

>less than 1% of manga in existence is being scanlated
>you can't find something else to translate
You sound like a pleb fuck

As a raw provider. That's why they exist

This is how you get bad quality releases. If you have one person who is decent at one or two things and they suck at everything else, then them doing all the work will result in a low quality release in most cases.

Imagine how much smaller it could be if there weren't jpeg artifacts fucking with the compression.

thirty-odd pages at x1600
raws are 12
finished is 18

Do you just rip tanks or do you archive chapter releases too?

I only archive chapters on a select few sites where it's the only place to get HQ versions of the images. Otherwise no point in chapters over tanks

Do any of you guys ever get bored with scanlation and try your hand at other things some times? Just for a change of pace.

I started doing subs for little things I translate, like this I thought about trying to make a fansub group. But really, the fan sub scene is dead these days.

either you snipe them or just don't scanlate anything at all sicne there's no point.

I subbed for a while. Then I forgot about it and now my raw provider's dead.

pottery

Attached: isekai.jpg (751x390, 77K)

I wondered about this and looked at some stuff
this is anjou(left) compared to Ookumo(right, below, 2500px horiz vs anjou's 3043, ookumo was downscaled a little to deal with screentone zoom issues)
I used ookumo since i know personally that it was cleaned (relative to anjou) well.
the linework seems to be impossible to fuck up, (and is a very strong argument for higher resolution regardless of cleaning ability) but if there's going to be any artifacts, that screentone on anjou would show it, but due to there being nothing to really compare it to it's really hard to say
I'm a little concerned about the small holes in the ink on Anjou's side, though. it might just be a analog vs digital drawing thing, but now i've gotten distracted with working on stuff so fug i can't really draw more conclusions

Attached: Untitled-1.png (869x762, 92K)

He said find good manga not find crap manga.

Any translators with some spare time?

You forgot the most important one: Never EVER read the comments on Mangadex.

Why?

I read them since some of them are funny. I see better discussion on other sites.

Because I need a drawing done, obviously

Because they're either blind praise or discouraging?

Drawing?

Corollary: never read the comments on sadpanda either.

Both, plus scanlators drama and god helps you if you're working on the most popular stuff because then you get nitpicked to death or people discuss the most pointless of things.

That's tips for being a scanlator, not a pussy-ass nigga.

Mangadex is autistic and God help you if your shit gets popular, but at least I've not seen anyone ERP and shit.

Where the acutal crap do people get the raws? I can only translate porn.

Just in case anyone doubts me that mangadex is mild as fuck, here's proof.

Attached: uh.png (734x489, 38K)

Wtf?

I wouldn't mind having a go at mrs Jones, if you know what I mean.

I already know that at least several of the readers for the series I translate speak Japanese, so luckily they've liked my translation choices so far.

Help, I accidentally hit some key in Photoshop and now it's like I'm shift scrolling permanently, what do I do to fix this?

Try Control+Alt+Delete to access task manager.

Install GIMP.

Install Affinity Photo.
No, seriously.

Well, restarting the program fixed it, but thanks for the laughs.

Where are you all sourcing raws? I've been reading via lhscans.net, but it's not great.

Sometimes, they make me chuckle. There's this one poster who quotes himself for some reason and I remember some dude from the batoto era who would always end his sentences with "mmmm" or something similar.

>don't discuss manga
then what's the point scanlating at all? you're doing it all for.. nothing? hahaha at least you can try scamming people with your sick "mom".

I actually prefer the comments on kissmanga and reddit. I tend to see more in depth discussion and predictions there.

Seriously? lhscans?

well for some series unless some kind user posts them they're not easy to find

The thing I hate most about Mangadex, besides the comments, is that any manga that deals with adult themes gets bombed with low scores. For example, the manga Holiday Love deals with themes of martial infidelity and because these incels are so triggered by the very topic itself, they bomb the manga with low scores.

>martial
marital

>caring about score
eh

>caring about anything
eh

Mangadex is full of beta faggots who hate NTR

Attached: 1541784701104.png (706x719, 406K)

>deals with themes of martial infidelity
Is that what they call cuck porn nowadays?

The guy cheats on his wife, incel.

Tell that to anyone with comcast.

So you're saying I should start translating a ton of NTR oneshots?

Downboated, was expecting cuck porn.

How's the quality of the images from amazon? I want to get the next issue of big gangan and I'm not sure if I should get it from amazon or rakuten.

Most likely identical. Gangan is locked to 1280px or some atrocity like that and has quite some artifacts.

yeah

Move aside brainlets, MATH coming through.
The argument for 1600p images is logically nonsensical, primarily because printers can print at any DPI they want.
There is no scientific basis for mastering images at 1600p.
By the Nyquist-Shannon sampling theorem, we need to sample at twice the frequency of the maximum resolution of what we are trying to record.
Humans hear at a frequency of 20KHz, so the sampling rate of audio must be greater than 44KHz in order to not lose any information.
If a print shop prints at 300DPI, then we must scan at greater than 600DPI in order to not lose any information.

From the perspective that a scanlator is an archivist, they must scan and create masters at twice the printed DPI, whatever it might be.

From the perspective of people who serve images, you can simply serve at multiple resolutions... this is a solved problem.
developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images#Resolution_switching_Different_sizes

Is there a way to estimate dpi from a piece of paper?

You're the brainlet here, but I'm not typing up an explanation right now, because I'm on my phone.

Went for dinner. Didn't mean there are visible jpg artifacts (don't even read anjou) just that they mess with png's compression algorithm even at q100. I'm more familiar with both gif and dng/tif since I work with those more often but it really should be similar for png. Basically if the group went straight from psd to png8 they would be able to squeeze out more size than the current jpg>png>optimize that you guys are doing.

Do you have raws for Strange+?

Get a magnifying glass and count the number of dots you see in an inch.

That is a tricky problem, because even though printers might print at the highest resolution they can, artists may draw on a digital canvas that has lower resolution than the print shop.
In the end, the best you can do is perform an initial scan at the highest PPI you can scan at, which is probably 600PPI.
Scale down the image (i.e. to 300PPI) and then scale it back up (i.e. to 600PPI). If your PSNR score drops by a lot, then you have lost information.
A less rigorous, but probably easier way would just be to note how crisp the edges are before and after.
If the edges are blurrier, then you have lost information.

If you're going to make a completely worthless post, then shut the fuck up you phoneposting loser.

Another technique I should note, now that low-resolution digital releases are getting more common, is the use of deep learning super-resolution techniques such as Waifu2x.
There is nothing wrong with them, because from an information point of view, because where algorithmic techniques aren't able to create information out of nothing, machine learning techniques can "invent" information by predicting it using historical data.
You might as well ship such a trained model along with your source images, but I guess there's something to be said for the typesetter's half of the work.

The only good upscaler for manga is Waifu2x.

Attached: waifu2x.png (1300x1836, 380K)

it's the only one people use
and it's shit for anything that isn't simpleass digitally drawn shit
but people do it anyway

Some patterning in the tone was lost (?) in the third panel.

Anyone working on non-jap stuff?

Attached: download.jpg (269x188, 18K)

Do you have abything better?

You don't know shit about scanning.

Posted from my phone.

You don't know shit about information. Enjoy your blur.

yeah, nothing
or in a dire case, not upscaling 2x, but 1.25x or 1.5x

May as well use Topaz Gigapixel.

Don't fucking upscale?

That's the worst upscaling algorithm I've ever heard of.

It was a negligible amount. Here's a gif comparing the original to the downscaled upscale.

Attached: Comparison.gif (832x1008, 218K)

oh so the source was shit
carry on

Fucking hell someone else actually did this. When I was really desperate to read something I used Google translate from my phone. Tbh it works better than expected it's just annoying taking a picture from your phone.
I made the whole process easier by using nox (an android emulator) on my pc and just dumping all the raw pages onto the emu filesystem. The GUI wasn't great to use but it took about 5-10 seconds per page.

did you turn off dithering

why the fuck do you think people upscale

For what?

That looks like the same thing, just taught on photos rather than manga. I don't think it'd work very well.

Doing God's work user. How do you plan on distributing it? Just a torrent I hope?

because new scanlators often think that there's some kind of XxXSceneXxX cabal that controls releases and that they need to meet some autistic requirements designed by some retard in order to not get flamed

Waifu2x destroys screentones.

That's a shame.
It's probably because Waifu2x's dataset did not contain enough shoujo manga.

I think it's mostly the denoise function, but I've never really tested it.

when you made the gif

Attached: .png (244x208, 11K)

Comments on sadpanda are often pure kino.

Its still only half useful at best. I don't really recommend it to anyone.

Attached: Untitled.png (1761x1008, 543K)

The problem is if you turn off denoise it has nothing else going for it. It'd just resize it at that point, no?

No, I mean why would I?

Here's the same gif with no dither and 256 colors. See if anything changes for you.

Attached: Comparison2.gif (832x1008, 247K)

dude just use kanjitomo

No, it still upscales with the neural network magic, the denoising is its own thing.

share.hotmoe.com/
jpraws.net/
zipraw.com/

You just reminded me. There's no open source manga aggregator, is there?
Meaning, self-hosted.
One of these days I'd like IPFS to solve its vertical scaling issues so that I can keep an archive and not have to worry about distribution when other archivers die.

>From there I break the sentence apart and figure out each word
There's one massive glaring problem with your method that anyone who knows Japanese would catch in a second.

Japanese isn't English. Sentence structure is not the same. There are many cases in manga and doujin where a sentence is split between two text bubbles. It's literally impossible for ORC to account for that, and unless you know Japanese and what to look for, you're never going to know if a text bubble is a whole sentence or half of a sentence.

>IPFS
I've looked at it when someone mentioned it a while back and I don't really see the appeal. What does it do that torrents don't?

Waifu2x is the best at upscaling manga pages and taking off jpeg artifacts, but it doesn't do both together very well. Upscaling with no denoising, then using the denoised page to take the jpeg artifacts off is the best way to get good results, since there's no blurring and no noise at the end.

Attached: atsumare.png (1336x1920, 439K)

Did you screenshot a page to google translate or take the picture of the page itself?
Opem the full image in a browser or something and aline the camera to the bubble, never failed for me.

The main issue that torrents have is that it's possible to split the swarm by distributing the same files using different trackers, or even the same files in a slightly different archive (i.e. by adding .nfo trash).
IPFS gives each file a globally unique hash (as unique as hashes can be), which allows the global swarm to be able to share the same file, even if it's in a different location.
Essentially, it's what DHT was supposed to be, but never was.
But IPFS has vertical scaling issues--the same issues that any DHT comes across, which is that a large enough amount of small files will exceed the DHT's capacity to hold them all.
Which is why the Touhou Lossless Music Collection crashed so many torrent clients.

I've tried that, and it still turns out awful.

The page I posted was doing that, and it turned out pretty well.

I had the raw files copied over to the emulator. In the Google translate app you can load images from your phone. And I stopped using this method ages ago, it's retarded. I'll just learn Japanese at some point.

This is why I translate "loosely", so it looks like my exaggerated translations are on purpose. And like I said, I talk a while longer and look stuff up online to make sure I got the gist of what the sentence is trying to convey.

Does anyone have a ripper for shonen jump+/pocket magazine? I've been using a ghetto ripper that saves images from the canvas, and since the viewer doesn't degrade the unscrambled images it seems to work fine. But I'd like to get the pages in a more, well, proper way.

Attached: 魔都精兵のスレイブ_01.png (764x1200, 816K)

I don't just run everything solely through google, you know? If a sentence feels incomplete, I see what the next sentence says and then piece them together. I have a basic understanding of how Japanese sentence work since I've been doing this long enough.

I'll grant you that longer more complex sentences do give me more trouble, but I just go extra loose with these sentences and try to match the context of what's happening in the panels.

What ghetto ripper are you using?

Lingodeer plus human japanese

>feels
Real accurate translations you got there.
Tell me, what do you if the sentence in question is complete is complete, and you just can't tell because it's using syntax that mechanic translators can't handle well?

>i- it's loose!
Yeah, I'm sure your mom is, too. Doesn't make it good.

It's a greasemonkey script I made myself. It gets all the loaded images in the canvases, puts them in a zip, and saves the zip. Simple stuff.
pastebin.com/VuDAe6hB

I type it into collinsdictionary.com/us/translator
and then take my best guess.

I'm actually translating a pretty popular series right now and no one's caught on yet. I'm actually thinking about bailing with some excuse, since if I'm caught I think some people might legit want to kill me.

>translating with an identity
user scans will always be superior

thought I posted already but I switched back to this tab and the captcha errored

Attached: dither_toggle.webm (818x972, 272K)

Yes, since you can always say you just uploaded your guesswork machine translations as a joke and no one would take them seriously.

The odd thing is, I might just be lucky so far, since even the people with some Japanese knowledge haven't caught on to me yet.

Or, "the people with some Japanese knowledge" are dekinai jokes and are doing the same thing you are.

One guy commented on a particular character's speech pattern in the original Japanese. Then said he understood why I translated it differently than the previous translator did.

Something I grabbed some threads ago.

pastebin.com/NLnKGRAP

Honest question here. Would you rather someone do the work of doing a guesslation like I am for you, or would you rather have no translation and have to do the work of guesslation yourself if you really want to read the series?

When I read manga I typically fit-to-width on my 4k monitor. So 1600px releases generally look pretty soft to me.

Attached: hurp.jpg (3840x4320, 3.5M)

Easy answer. No "translation" at all.

Are you an ESL monkey? Then no.
Are you not an ESL monkey? It depends on raw availability, and whether it's popular enough that someone better might do it.

The series can't keep a translator for some reason and I'm a native English speaker. Alot of people say I'm better than the previous translator, since he made alot of questionable stylistic choices.

So basically, it's probably a case of my translation being less accurate, but sounding better.

I don't know what that is. A link would be nice.

I don't really know. I wish there was a way to distribute the raws safely and conveniently, I already have so much work to do acquiring and organizing the raws I don't have time to also make a distribution platform.

the 1600px release has clearly lost information.
archivists roll in their graves.

Thanks Ill have a look and see if I can make it work for manga-bang. Its a good excuse to mess around with js at least.

I should probably add that I got A's in my creative writing classes in college, so my writing ability likely helps mask my lack of translation ability.

yeah, there's soooooo much stuff out there that's cool but no one translates, and there's sooooooo little TL talent out there to do it.
At a certain point it's worth doing guesswork just to release stuff, if for no other reason other than to give more access to the comic's art.

>popular
isekai

Writing ability correlates directly to translating ability.
It's a literal difference of between being able to express something and being unable to express something.

oh yeah i was gonna cough romcom hero and the demon king cough
since the dude dropped it due to low confidence in his TL skills when I just wanted to see the titty massaging extras

Yes, but can you guess who I am?

Maouyuu?

It's simple, I just have to find the translator who isn't clearly a SEAmonkey.

I don't read isekai shit.
slime

Nope.

ideally they would be from the same master but those are from two different releases presumably from two different masters but that was as apples-to-apples as I had on hand. I'm sure someone has a better example they can muster.

I'm just not convinced 1600px is all the resolution you need because in my experience it comes out a little soft. Then again I'm a tiny niche who reads manga fit-to-width on a 4k so I don't expect people to cater to me I just generally like to have the option of higher resolution scans and if given that option I always opt for the highest available.

I've always wondered why mangadex doesn't do that. It serves big-ass JPEGs to shitty phones instead of the moderately compressed WEBP images that they should get. Mangadex always has issues with bandwidth.

You could have used Google to check the internet plans and pricing on European websites instead of shitting up this thread.

It's likely because they are picking up from the grave that Batoto left behind.
Batoto's claim to fame was that they don't mess with your images. Everyone else would add watermarks and compress them to shit.
Whereas twitter's research has determined that phoneposting scum doesn't care even if you send them a 60% optimized JPG.

There's also a likely issue of storage. Batoto at the time of death had an 8TB image database, which has surely only grown bigger. With parity and CDN costs, I'm not really sure they were able to justify to themselves that they could experiment with multiple sizes and formats of images.

mangadex has grown way past the size of batoto though. They're forking over almost 700 bucks per month for their servers. I don't know jackshit about this but I'd assume that they need those in order to meet bandwidth demand. A few TB of storage can't be that expensive compared to that.

They're taking way too much.
A website for English scanlation is enough, but they want way more than that it's so stupid.

depends on if they can keep funding up. They talk a lot about wanting to kill adware infested aggregators and they're certainly stealing a lot of traffic from them. Initially i just used it because uploading was easy and it spread from there but a shit ton of people seem to be reading on mangadex nowadays

Twitter is absolutely right, virtually no point in sending large images when 99% of the users aren't going to zoom in to get more detail. And honestly I doubt phoneposters even want high detail shit since that'll shit on their bandwidth.

A good clean site with good quality scans always wins out. Anything that's not clearly designed to extra max ad revenue and using the shitty prozess reader is auto 10 miles ahead of every other reader.

Actually, I think that there is a better technology that we can use, which is present in the H.264 video codec. They call it "Scalable Video Coding" which implies that you can use multiple parallel streams, which when combined, can produce a higher resolution stream--essentially allowing you to encode multiple resolutions into the same image.
If both the client and server understand the protocol, you can send only some streams instead of all of them, which can save you bandwidth.

I believe PNG has something similar. Interlacing. But it has a different intention.

Chink sites like kiss didn't do themselves a favor when they started cramming the site with malware infested ads, then they have the audacity to ban ips who use the site with even a modicum of adblocking.

They were stupid as fuck if they thought people would turn off Adblock and get their computers infected. People did the logical thing, they searched for something safer.

sure, but it was always like that. For some reason batoto didn't gain remotely as much traction as mangadex

You are making quite a lot of assumptions about the screen being used, the preferred method of reading and I guess the "platform."
1920x1080 is not an ideal resolution for reading manga. If you do insist on reading at this resolution then do you want to be scrolling? if so there's the Korean style whole chapter in one very long page deal you could be using (either by scanlators using a single image or by using a reader that displays separate images as such). In this case you want the height of one frame to be 1080. I think most manga pages are roughly 3 rows of 2 which means 3240 is where you would be before zooming out.
A much better method imo is to go 1080x1920 at which point you want exactly that resolution for a portrait page.
Phones, tablets, etc are usually used in portrait mode but they can have much higher resolutions. Monitors are also improving so up to 2160x3840 should be the goal.
As for "platform," if you are using an online reader then higher resolution jpegs downscaled by the browser may be better than a lossless image at the correct resolution.
I don't get the 1600 format. Is it designed for portrait with website banner / ad space? fuck that.

batoto
>ugly website layout
>register-wall to read anything
>no long strip option
>no comment section / chapter
>sjw infested

>I considered going around 2050, but I think it might make it hard to read on such a small screen
Why are you not scaling the text with the resolution? you want cleaner crisper text too.

Kiss started aggressively blocking adlocking in the last couple of months. There is no coincidence their traffic dropped precipitously the same times mangabey exploded in popularity.

What do you think these people are trying to destroy mangadex by DDoS attacks and making "clone" sites? €£¥/Money, they are not making as much of it as they used to.

It's an open secret, Chinese manga sites sees mangadex as public enemy number 1.

I don't think that's bad or that we're close to that. Physical copies are far more detailed that what we see on screen.

I've had mixed results reading Chinese off of Taiwanese whiskey bottles. I suspect the issues are more the rounded surface and the refracted light coming through the bottle. Give it a go man.

>You could have used Google to check the internet plans and pricing on European websites instead of shitting up this thread.
Or you could not shitpost this thread with your random, unprompted seething about americans. on an american website. And yet you continue to do so right now. Fuck off, retard.

Releases with higher than 1600px in height are really nice for editing and fanart coloring
I do believe that capping releases at 150dpi for common consumption, and 300-600dpi for HQ releases intended for fanart and archiving, would be an ok way to save resources
Though there's plenty of releases where the extra resolution it's unnecessary, specially those made from terrible digital RAWs, but resampling on a case by case basis requires quite some effort and time
That's among the most efficient lossless optimizers for PNG
Still, lossless WebP it's usually faster at encoding and compresses better, no reason to stick to the past so much
And for most people mozjpeg at ~80% quality does great, even with line art, their deringing/overshoot stuff really fixes the most glaring issues JPEG has with lines, but pngquant with really few colours would do even better, specially on pages with simple shading and no colors

Why are you conceding? You were right.
The rest of the screen will be full looking like shit, but it always looked like shit. You didn't make it worse. And you can zoom out to go back to what you had in the first place.
If you do not intend on zooming out on image that looks bad and use your normal resolution settings like a normal fucking person then the clearer text is much appreciated.

Attached: Untitled.png (548x559, 16K)

Turn off cleartype before doing this comparison. You don't typeset with subpixel rendering.

Attached: Untitled.png (258x128, 4K)

So it's alright to being alright from a technical standpoint and back to the question if clean text in pixelated art is alright. To which I really don't know the answer.

It is in no way worse.

Attached: Untitled.png (1261x810, 314K)

Kill yourself

Looks bad but that may just be the font. I guess I'll do both versions of at least some pages of the next chapter I'm typesetting to see what it looks like and ask some people what they prefer

How's your interest in scanlation holding up? I'm getting pretty weary of it and I don't even do much.

Attached: let go.jpg (395x395, 13K)

I clean and typeset one chapter and I can't bring up the motivation to do a second for a month
translating is whatever though
when will someone eliminate the bitchwork of cleaning?

Pretty weak at the moment. I have a few chapters I have to finish cleaning, but once I catch up hopefully I'll be more motivated when the next chapter comes out.

I've been working on one doujin for a month now, I only have 7 pages left, but the redraws are killing me, so it'll probably take another month to even clean.

Anyone ripping the final volume of Kuro?

Though uasaha quit but they rebranded as rurika and danke has mentioned he has no interest in the series. Dunno if the other rippers care since I don't lurk the wino threads.

I'll probably have all the new chapters of Shadow House translated by the start of June.
The "people" who comment on Mangadex are absolutely subhuman, when half the reason you're translating something is because you'd like to be able to discuss it with other people reading their stupid jokes and retarded plot mispredictions is downright disheartening.

Attached: sh.png (794x766, 419K)

I've been curious about that series. I'll probably binge it once you catch the series up.

I like theorizing as long as they can back it up with evidence and not shit like "Well, this female character was gone for a while, so I assume she's pregnant now for some unexplained reason."

It's not as good as Kuro, or some their doujin, but it's a fairly interesting series.
There's nothing wrong with theorizing in and of itself, that can be plenty fun, the issue is the baseless or otherwise stupid shit.

>You're going to need more than one.
Would you seriously consider sharing them were I to provide long-term hosting/distribution? I could work with you to either Syncthing the contents to a dedi or perhaps physically ship drives.

>Japanese company looking for jp-en translators for games
>with 1 year experience
Does anyone have an idea how companies feel about scanlating as experience? I'd imagine Japanese ones would be more prickly about it

I'm trying to get into scanlation but I can't seem to get into typesetting. All these time I always translate and then ask my underclassmen that still haven't graduate uni or my co-worker but most of the time they're busy with their project. I tried following a guide but I can't find the right font to use, can someone link me to a good font pack?

Used to make 1 chapter/week, did nothing for 3 weeks, nobody noticed. All's good I guess

lmao just use wild words and call it a day

Okay user, thanks for the suggestion.

>I'll probably have all the new chapters of Shadow House translated by the start of June.
Nice, guess I'll look for scripts to do a few chapters.

>in their defense JPG does usually do smaller filesizes in most other usecases
None of those usecases are black-and-white images that alternate between a million tiny details and big flat areas.

>Save for Web with 256 colors
Nigga there are a total of 2 colors.

13dl

As long as you stick to isekai I have no complaints.

Buy your own manga and scan it.

I only have to do 1 chapter per month so it's all good. I wanna start on this doujin too but the redraws make me discouraged.