General Discussion (previously Canon Question)

The short answer to your question is about the doujins is yes they are canon. They continue the plot in a non-animated form. Some of the more “explicit” material in those doujins aren’t so much canon as much as what they are at face value. The Shin Tenchi novels are also canon and show events that happened prior to OVA 1. GXP and War on Geminar both continue the TMR! plot line. Gxp takes place after OVA 3 aka “Ryo-Ohki” and War on Geminar takes place roughly 15 years after GXP.

Also Isekai no Sekishi Monogatari is the Japanese name for War on Geminar.

Dual does not directly tie in with Tenchi but it does “connect at the corners” according to Kajishima the creator of Tenchi and Dual. Dual does make references to Tenchi however (if you have seen the GXP episode called “Parallel” you will get the connection).

Hope that helped!

It also doesn’t nothing at all for handwritten font, and currently the better method I have does horizontal text only, so character have to be chopped and reoriented from vertical (which there’s a loooot of).

Images with heavy background art too, are neigh impossible… you could probably clean the page by erasing / painting the unwanted art leaving the characters, it’s just a huge time sink.

We can talk if you want to do something… just we really truly need to get a proofreader if we’re going to do it right. Machine translation only gets you so far. Your credibility is shot when you translate something wrong, and it makes it to the news…

If you review the translation section here, I should have some guides as to how I’m doing it currently. CamDictionary is the best thus far. KanjiOCR is the older way.

Your eyeball may be able to look at something and think it’s great, but a 3 MP camera doesn’t like it nearly as much.

I did just play around with some manga and doujin, and yes even some novel scans, and the results were mixed. I did find out that the google translate app will do vertical text (small woot) but getting a good capture for it to focus on is hard. Just about need a tripod.

Novels, I don’t even know where to start there. Nearly every line has sort of a subscript of characters, which is probably a modifier to what it’s attached to. Honestly I don’t feel we’re setup to do novels, those are massive amounts of text, surely complex word usage, and the scans are so-so. About the only benefit there is it’s all machine text.

The larger a block of text, the more a machine is going to butcher it. Shorter one-liners or sentences are far more accurate.

Chunks of manga or doujins, maybe interviews… is probably the best to hope for with what resources we have at hand now. Sure if someone can get a translation group to pick up something… more power to ya. I’ve tried here or there. The common issue I had is that it was so old… Almael would be the one to really have an inside view on this as he’s worked with baka-tsuki.

Kanji OCR is older than dirt, and requires a lot of photoshopping of the image first, and generally a couple of passes before you’ll get the text ripped.

Apps are so far quite a bit faster at the job.

Happy Birthday JGZinv.

Well, where to begin?

It’s nice to have a team but when it comes down to old and no-longer popular titles you are likely left alone. I did have a dozen people over the years but sadly they all jumped off very soon. Recruiting is very difficult and bleak.

What you need are dedicated people willing to work it. It doesn’t matter how good they are as long as they do some work. Without dedication and commitment, forget it.

As to translation it’s not impossible to do it with machine translation but you need to learn a bit about the language. At least know roughly the kanas, the structure of the language, have some dictionaries at hand, a Kanji search engine, and a romanization engine. You should learn to recognize them after working for some time… You gotta look at the romanization rules; develope a feel for engrish. The japanese are very “flexible” here. You will always end up making a judgment call due to the fuzzyness, possible variations, typos, etc. Depending on the author’s style/fancyness half of the time it could be very easy or hard to decide… Tenchi is probably way easier than Seikai or KOR, I doubt it has multiple languages to deal with.

If you got the ocr version it’s more likely to get people to translate…but don’t get your hopes up.

Imho, Finereader and its japanese Yomkaku are the best programs. I use Yomikaku but it’s way too cumbersome if you don’t know japanese. I dunno about the new Finereader versions but it should do very well. I used to use it before. All programs have difficulties with changing background and colors so there are times when you gotta invest in preparing the images for OCR. Depending on image format or sources there are other cleanup tricks … often very complicated; too complicated to explain here.

For me the ocr version has a lot of advantages, especially with the Baronh. Search and replace makes it very easy. Of course you gotta have special markers to prevent replacing things you don’t want. Makes it easy to do Baronh corrections. Well, JGZInv got my detailed guides…

The perhaps most important trick with tranalation machine is to get rid of names/terms/places by translating them yourself first. Search and replace. That saves most of (google) machine confusion.

The japanese machines are smarter and force you to do it yourself anyway.

Backtrack sentence structure, japanese is reversed, and typical for asians compact and to the point.

Often connecting sub-sentences with the same “thing”(verbs) in mind… In a western language we would use a sentence for each part.

I’m going to snag a copy of finereader ans see how well it does. I know it’s been around a long time.

I tried all the programs you mentioned strat a long time ago, and they didn’t work any better then either. Readiris is a joke, as are the rest.

The issue though, is scans. IT can do a very good job even against a background image,

however you do need scanner quality scans. The stuff shades has at G.hentai is not sufficient

for good rips, it’s about 30-40% smaller than would be good.

We still desperately need a Japanese proofreader… but I think with the tech we have at hand, we can

definitely get OCR rips.

Page 20 is close, like you said, but with the kanji mostly eliminated, other than just stripping

the rest of the text for the sake of doing it, not much we can do with this.

162 is usable.

BTW I’m a PS CS3 user, so no worries on that, I’ve done all that before too.

120 looks like mostly Kajishima complaining up a storm about getting Tenchi through the production

process. *cough*sounds like lucas*cough*

I had a full set of Nontans once upon a time. Shades and ANB got most of them.

I secured a source for novels… OCR is mostly ok… the subscript like characters are throwing things for a loop though… arrgh.

Hasegawa’s 1st novel title is something like…

Tenchi Muyo! Ryo-Ohki

Successive Visitors: A tangled web of love!

(It should be more romantic than that… the book deals with new years in some way from what tiny tiny way I’ve looked at it so far… but I can’t seem to think of anything better.)

and a couple of the old thicker ones.

Doujins I think will be the easier ones to tackle due to the text. They’re also shorter, so we can

get something done, and the 2 Heisei-bans are partially Dual related, so I’m most familiar with the

subject. They do have Tenchi in there too.

The Kanas/Kanjis should have at least a size of 28×28 preferable more like 32×32 or better.

I usually scan and save as bmp first.

Then I use Irfanview to reduce the colors. Do not reduce colors in one go! It’s best to do it step by step to preserve the structure: e.g. to 65536 to 256 (to 16 (to 2)). You may try and skip a step if it’s possible.

For a reduction to 256 I use the ‘radical image optimization tool’ plugin, it usually gives the best results. File->save for web(plugin)

With Irfanview you can batch every work just make sure of the settings each time, because the last setting might be different. You can save batch scripts, too.

Then save it as bmp or gif then batch convert to png gives the smallest size with best quality. Saving directly to png does not always gives a smaller size due to often bad implementation of the format. PNG can store lower bit colors but people usually program it to store 24 bit so it doesn’t always gives a really smaller size. It’s bothersome when you do some 3D mod and 65552 colors become preferable.

Jpg can get smaller but it blurres everything and may introduce more smeared colors. Well, just saying.

This wasn’t posted in here, as we did it over minchat, but with some aid from crazed I compiled a big substitution file dictionary for the translation aggregator. Basically kanji names are a great way of messing up the translation, so this swaps out anything it has an entry for. So far, been working pretty well on Dual stuff. I’m adding a line here or there for stuff I recognize but didn’t have quite right.

Translating this stuff is scary….

I’m not sure if this is Kajishima trolling the fans… or literally his intention for the end of Ep 14, but it starts in on Kazuki and Mitsuki R having a baby girl as a result of Ayuko’s deception. Yayoi’s the step parent, the others act similar to Tenchi, in taking care of the baby together and trying to figure out how to change a diaper. Reika gets angry while Ayuko just laughs with a “Hohohoho…!”

Originally MitsukiS was going to be more towards Yayoi’s role. Dee would have been older. Two other new characters would have come in and been much more like NGE. There was a blond haired spy woman that infiltrated the EDF during their acceptance of more pilots from around the world.

*shivers* Take it away… I don’t want to read it anymore…!

*Runs off into the night*

I’ll try to get a program set together of what I’m using. Bundle that up.

Scans I’ll have to organize, the doujins are all in one folder, far as Dual goes, but I’m not sure it’s complete.

A note towards strat_guy, is that I feel we need to take a good look at some of the English

translations we’re doing as a quality check. Some of it just doesn’t sound quite right.

I know I’m not able to check it all everyday, but I’m trying to get around to it.

I’m also probably going to stop trying to translate the sidebar text from A!MG, the stuff is just too error prone with the quality of scans we get. Granted it’d be a leg up over what anyone else has done but it’s taking 8-10 times longer per page to get a rip. I could finish all of the remaining 20 pages roughly in a handful of hours, or spend another 3-5 days on it.

The early bird gets the worm in this business so to speak, this needs to get out there so we’re visible. Plus it’s the long awaited 25th anniversary 300th chapter. So we need to get it done, with our name on there.

So I’ll probably be switching gears with that. Otherwise, I’ll try to go back and proofread any pages done thus far…. probably need a new “tag” of some sort for the title.

I might also talk to dagon at some point about the finished results and how we’ll organize the subforums, since we’re really close on the Heisei-ban 2.

I’ll start cleaning AMG since Strat has HB 2.

AMG has already been released somewhat privately and incomplete I might add

as a commission by someone. I’d prefer we show ourselves of higher quality, so

I’ll still work on releasing it.

seems to be picking up the substitutions from that list we made.

Just need to clean up the translation results and stick text on pages.

Still working on the final editing of AMG 300… getting there.

and not staying up to work on the ch. 300 editing.

So… see you tomorrow.

Not sure when I’ll be caught up still.

so I’ll see if combined with glass or some other elements if that isn’t a decent means to get some of

our more rare things digitized.

I think the trick is more so going to be taking a sheet of glass, perhaps a mirror, and getting it angled right so

you can get a scan.

What really befuddles me, is the DIY book scanners that take a few hundred pages an hour, use(d) mid to low range megapixel cameras. Those were high cost parts of the assembly at the time. Scanner wands have been around quite a while and the price has been in the $80 to $300 range, and you’re getting a better DPI.

Can’t say I’m going to reinvent the wheel, but maybe the books scanner concept needs an overhaul.

As it is now, I wouldn’t use a glass to smoosh the books out, the wands don’t take photos, they scan (no shit), I have a feeling they are design to touch the page.

It might take super long, but if you scan page by page, getting as close as possible to the spin, you could have better results.

Seriously though, for the doujins, I don’t even understand why you would need a glass. Thin doujins are the closest thing we have to a magazine.

The thick ones however have spines and are very stiff and easy to crack.

What I’d like to find is some sort of glass or “view shifter” that is somewhat like a periscope.

Where I could place the flat glass shifter, and it would fit into the book so the whole page was covered, then use the

wand scanner an inch or two away from the spine on the glass and get a proper full page image. Some things you just can’t

scan well without destroying it, so that’s what I need to figure out.

I got the scanner wand in today. Going to start messing with it in a few minutes.

The two issues I see with it is there’s about an inch of housing to either side of the wand (not edge to edge) that keeps

it from being close to the spine.

The second is that it has no real ability to scan hovering above a page or not directly on the page. The reason is two fold.

A. The rollers on this particular device actually activate a burst light when it moves. No motion = no light.

B. When not directly on the page, the image is blurry and smeared all to hell.

So my next course of action is to get some glass and maybe a mirror, see what can be done with that.

Otherwise, I’ve only messed with JPEG at medium (600dpi) and hi (1050dpi) and the scans are usable.

There’s a bit of a crosshatch or almost like a texture effect in the solid areas of color if you zoom into the resulting

pictures, but hi level scans at 8MB for a book cover are really nothing, and scaling down the image for web removes

that anyway. PDF might be different, haven’t tried. Scan speed I’m going slow (to me), but I’m faster than a flatbed, haven’t had a smeary image yet when I was using it properly.

Color reproduction is fair to good. Certainly web usable, but I wouldn’t use if for color print reproductions without some tweaking against the source. The same could be said of flatbed scanners, albeit to a slight less degree.

For the $42 I paid for a $80-120 stick, I think it’s worth it for the functionality it offers. It’s a one trick pony, but that’s all

it ever promised to begin with. Supposedly I can get several hundred to a thousand scans on a 1 GB microSD card, and the built in battery will last a few hundred on it’s own. Don’t know if the 1.3 tft screen is worth it, probably not, you really can’t see squat for the previews on it. The extra 140dpi in high mode over the next model down is probably kinda sorta worth it though.

It does offer a scan direct to computer mode, as well as ABBYY FineReader 10 (should sound familiar to some of you), and a cloud app if you want to do that. Otherwise PC hookup just sees the memory card as flash storage and you can pick files right off the card.

Edit – Trying it with some heavy plate glass, just produced non-useful mostly blurry images. Probably find some thinner scanner bed glass before I call it pointless, but not looking good.

The scans started out as 5200×5200 or so. Not sure what else you could want from it.

The source just needs to be flat, even staples in the middle of the page can be a problem.

The more annoying thing for me is I can’t seem to keep it from cutting off the end of the page for

most any given scan, or because you drop off the page the scanner shifts in hand and you end up smearing it.

It does work, just it has some quirks.