Superdoll feasibility

[DaFool]

I am gauging interest in an advanced paperdoll system I call Superdoll.

Currently we only have 2 paperdolls - Alice and Betty, each with limited expressions and accessories. We also have 2 paperdoll-like image packs (or layered .psd compilations) in Community Contributions thread which are more advanced in expressions.

What I find lacking is the amount of variability in these dolls. If you have seen one of them used at all, you have seen it already. So to have something different, you have to make a new paperdoll.

This is something I have to undertake for my own projects since the image requirements are ridiculously insane and to achieve a cinematic effect there should be enough variety in character poses and expressions. Thus I want a Superdoll system to:

- output multiple characters, each with their own proportions, clothes, hair, and accessories
- have these multiple characters share the same but very extensive set of postures and expressions

The number of combinations I envision to explode to almost an infinite number of possibilities. If designed well, no two Superdolls will be the same!

Now for this to be a real tool instead of just a collection of images I obviously need an engine. I'll be happy to use Alice 1.1 in the meantime, but I was thinking of *gasp* making the tool run on Ren'Py. If so, I'll need to know how Ren'Py can be used to output a composited image file with specific dimensions.

In essence, this is similar to that Poser anime doll (her name was Aiko or something, but I never tried Poser so I don't know), but I am among many others who prefer the 2D look (but hate the look of Flash characters), so I was thinking of doing this.

Opinions are needed, especially to technical considerations. I'm banging my head constantly considering the aesthetic aspects, but I think it might be doable, especially with using multiple layers.

I was also of extending this to include a Backgrounds Generator, but instead of Terragen (which is a random 3D landscape generator), this will be a 2D buildings and furniture layout generator, just using a lot of layered elements.

[Jake]

Um... I want to vote for both options at the same time...

I'm certainly more than interested in such a system, but a large part of my interest is in the way you intend to carry it out. The only way I can comprehend of executing such an idea would be to literally draw out each of the different poses and proportions for each article of clothing, each accessory, carefully set positioning info and so on... to the point that to include a new item one would have to do tens of drawings for that single item - jumper on thin short female with small breasts with arms at sides, jumper on thin short female with large breasts with arms at sides, jumper on thin short male with arms at sides, jumper on thin medium-heigh female with small breasts with arms at sides... etc. I'd worry that such a project has too many potential showstoppers; for example, if one got a significant amount of art prepared and realised that a pose was missing, it would then require additions to several garments, in several poses, in several proportions... too many chances to get put off and stop work.

Another option I can think of is a 'boned' or 'jointed' system, where a jumper is defined of in pieces - one draws and sets position data for 'jumper forearm', 'jumper upper arm', 'jumper torso female', 'jumper torso male' and then stretches and rotates and composites these images according to a particular scheme for each pose, but I suspect that wouldn't look too good in the majority of cases - if nothing else it would be impossible to shade and highlight the individual pieces properly, the result could look like it was being lit from a different angle for different parts, and things like outlining around the joints would probably look odd.

This is not to mention that it would be very difficult to coordinate art between more than one contributor to look consistent enough to mix and match - most people have a hard time accurately mimicking another's style.


So... yeah. How are you planning on actually doing it? How does it work? I'm certainly interested, and would be more than happy to contribute to such a project if I could, but without more information it sounds like either one of those "it would be really nice to have..." projects that hasn't been thought through much more than that, or a huge enough undertaking that it would be hard to finish. I would suggest that the art generation part is probably the hard part, though, and getting a program put together to do the compositing is likely trivial in comparison...

[Edit: The main thing I find lacking with existing paper-dolls that I've seen is less the flexibility, and more that they're nearly all dead-face-on to the viewer. Sure, most people do stand face-on when they're standing around talking to people, but we also get to see more of them than just that in real life, we're not limited to the one 'sprite', and three-quarters angles are usually much more characterful...]

[DaFool]

Regardless of there is general community interest, I'll have to create one anyway since there's just no way otherwise of outputting a lot of drawings.

Some things to note:

- This will be simply linework, at first at least. We'll worry about coloring later, and it will actually be better for people to color themselves.
- Definitely 3/4 view, side view, or what have you. Like I said, really going for cinematic poses here.
- Definitely hinges on a successful design, since I definitely don't want things to look papercut-like or like Flash pieces. But do note that even with traditional non-Flash 2D animation, each moving object - a hand, a sleeve, a foot, is drawn on a separate paper layer. So this will definitely apply here.

[EwanG]

I think even having a library of PSD files where there is a file of layers called "clothes", and a file of layers called "torsos", etc. would be a major help.

Of course, if you want to help out the hentai community, it would be good if the torsos had to be dressed

Assuming you had some files like that, I would think that RenPy could take advantage of the routines that Gimp uses for working with PSD files to then grab the layers that were specified by the gui and output either a custom PSD with all the variations you need for the character you chose, or a group of PNGs.

Just my off the cuff thought on how I'd do it as the apocryphal lazy programmer

[PyTom]

Now for this to be a real tool instead of just a collection of images I obviously need an engine. I'll be happy to use Alice 1.1 in the meantime, but I was thinking of *gasp* making the tool run on Ren'Py. If so, I'll need to know how Ren'Py can be used to output a composited image file with specific dimensions.

You can't. Ren'Py isn't set up for this sort of thing.

[DaFool]

Hmmm...

A really tacky method would be to imagemap the selections chosen by the user for compositing, then display the picture and then declare:

"If you like what you see you can click the center mouse button to remove this text box then hit print screen to capture the image"

but so much for transparencies, lol.

What suggested methods are there from separating the bulk of image layers (the "bin") from only those which the user has chosen to use (and hopefully have the ability to rename)?

I'm already thinking about rendering entire sets of images, I should just think about the image layers first. But I don't want to just have one big psd file.

The Alice program is fine for now, just need to make multiple dolls then.

[monele]

I'm really pessimistic about this ô_o... Especially if you're going for multiple poses and ... phenotypes (not sure of the word : "fat, thin, normal"). Basically, while the number of combinations would be incredible, you'll still an equally incredible amount of work to get everything in place.

Let's see... One thing that might help in the theoretical department would be to check out the Tokimeki Memorial Online chara maker. Has anyone ever tried this? If you check the game trailer, you'll see part of it and it's quite incredible. But I think what they did was make a *few* stand up dolls (maybe... girl/boy, small/normal/tall) and then make thouuuusands of accessories and clothes. That's the part where it hurts ^^;.
While combinations are usually numerous, keep in mind that once you've seen that particular shirt, any combination including it won't look brand new. In a way, the real number of neat combinations possible is the number of total accessories divided by an average number of accessories per character.
One thing that could help push the number up is through colors, but it'll still look a bit dated quickly enough :/

PyTom also pointed out the problem of styles, already seen in Alice : parts made by one person won't be compatible with those made by another person. You'd automatically notice the fact they're not made the same way.

While I love the idea, I think there is a problem ^^;

[monele]

Hey, there is actually a flash demo version of the chara maker ^o^. Have fun :

http://www.tokimemo-online.konami.jp/game/chara.html

You can check the video (first picture) or try it out (big green button)

Basically, you can change the skin color (three tones here), the hair color (4), the eyes shape (big cute or narrow mean), you can add glasses or moles, change the hair shape (front & back), the uniform (4 types), the pose (3 per gender), the eye color (3).
Of course, the real big thing is accessories but they give you only a few. Headbands, earrings, ties, bracelets, hairpins, etc... could really add a unique touch to each character, for sure.
I've seen taller characters on screenshots but I bet they're NPCs ^^;.

Note that 3 poses x 4 uniforms x 3 skin tones = 36 versions already for the body only ^^;... If colors could be dealt with some recolor function, it could already help, but clothes+poses is still a biggie.

[DaFool]

Thanks, monele, that was quite a bit of little fun

Something I noticed though, the posture + body type is the same sprite. For example, the crossed-arms posture is associated with the fattest body type. This is definitely something I want to avoid, even if it means in this example 9 postures (3 (thin, normal, fat) x 3 (one hand in front, both hands behind, crossed arms))

But the difference is that the dolls here are constrained to be compatible with the particular game universe. A truly useful doll will not be constrained to work only with a particular game universe (usually one that matches the artists' style). To combat the issue of it being hard for someone else to add accessories in a matching style...

1.) Uncolored simple mechanical linework. Much style is brought in coloring and shading. Reduce that to mere lines, and pictures will be more compatible. Then the user can color the sprites to match his / her game universe requirements.

I can take advantage of linear ambiguity with linework. For example, a line that might be the edge of one piece of clothing can become only a crease in another piece. A wrinkle on someones face can also be interpreted as a loose strand of hair. That means layers will not be strictly solid-object-based, which is what happens with colored paperdolls.

2.) Keeping designs simple and efficient yet effective. This is what they do in eroge->anime conversions all the time. They don't need to be full details, just the minimum necessary to make the characters recognizable. I am sort of going to quote Madarame in Genshiken on this...what is the minimum number of lines in a female character's drawing that is sufficient to turn you on? That's the holy grail I'm going for.

3.) Include as many accessories as possible in the default package. From here it might be easier for someone to follow the general pattern of shapes and objects to make a good estimate for future accessories. Groups of accessories that relate to a specific theme can be packaged together, for example, School Life or Space Sci-Fi.

This is all talk for now, but I really need a better way then doing sketches again and again, when they rarely look the same. I tend to be the person who works on perfecting one drawing while leaving the rest in shambles. So perfecting one element at a time then reusing is a very appealing idea.

In the end, in any form of media, its the story -- the writing -- that counts, everything else is second fiddle. If this finally gives writers but non-artists an unfair advantage, if that will produce only better and better stories, that's a good thing.

[Jake]

1.) Uncolored simple mechanical linework.

Leaving the sprites uncoloured is probably a good idea, since it will leave another element to be openly modified to distinguish one use of the doll from another.

On the other hand, you would have trouble from a technical standpoint if you wanted to provide just lineart on a transparent layer, since you'd end up having to cut lines away behind garments, and know when and how to do so. For example, you have a figure drawn in their underwear, and want to give them a jumper - you have to remove all the lines for their torso and arms, so they don't show through the jumper, and you need to know from the jumper sprite which bits to erase despite it having a transparent inner as well as outer background.

(One solution to this would be to render all the component sprites with pure-white internal colour and transparent backgrounds, so that they blocked out the lines behind them... but then this would make it harder to colour in the resulting sprite.)

1.) Much style is brought in coloring and shading. Reduce that to mere lines, and pictures will be more compatible.

I disagree. Well, partially, at least. I think you'll find that with comic art, more style is typically inherent in the line art than the shading/colouring. Take a look at X-Men, Love Hina and Fist of the North Star - they're all totally different in style, even once you remove any colouring or shading.

Even amongst my own art the lines differ significantly depending on how I go about it; if I draw in Photoshop, I tend to use a 1-pixel brush at close to display resolution, and end up with a broadly even line; if I draw on paper with a dip pen, I end up with lines wildy varying in thickness that don't look consistent with the computer-produced art at all. (Hell, I know at least one artist who doesn't draw line art at all, he blocks in black then paints white over the top to form shapes, and still ends up with something distinctly manga-style - and distinctly different - at the end of it.

I can take advantage of linear ambiguity with linework. For example, a line that might be the edge of one piece of clothing can become only a crease in another piece. A wrinkle on someones face can also be interpreted as a loose strand of hair.

I think you'll find that - especially with minimalist line work - this sort of thing looks a lot more awkward than you might think... :/

[DaFool]

Even amongst my own art the lines differ significantly depending on how I go about it; if I draw in Photoshop, I tend to use a 1-pixel brush at close to display resolution, and end up with a broadly even line; if I draw on paper with a dip pen, I end up with lines wildy varying in thickness that don't look consistent with the computer-produced art at all.

Good points. I don't like linework produced purely by tablet, so I'm debating between hand-inked and traced-paths. Traced paths will definitely be easier in flood-fill coloring, but there is a certain soul in pure hand-inked which is lost when the lines are perfected. But then again, once people start to color, it will become their arts, not mine, so while I want to retain the soul of hand-inked on mine, maybe other people would prefer the more crispy anime-two-toned CG.

EDIT: Actually, wait a minute, this gives me an idea. We could have different linework styles while still using the same positioning.

Example:
A. purely hand-inked
B. traced-path and touched-up / clean style

OR
A. Shounen (i.e. spikiest hair)
B. Shoujo / Josei (i.e. biggest eyelashes)
C. Seinen (Suzumiya Haruhi...well, maybe something else)

[monele]

Just a question : *who* is going to do all this? ô_o... That's an incredible amount of work.

[Choark]

This is a really nice idea, espically for those who have little to no artistic leanings and want to show there written work without having to wait for those damn lazy artists (;_; I'm sorry) and if you manage to pull it off more power to you, but from my perpective I actually hope it doesn't get used too much when making visual novels around here.

To explain:
For me a lot of the allure for visual novels is the artwork in it. I know everyone has there own personal peference on how much its the story and how much its the art, but for me I do tend to look a lot into the graphics. This is even true for the games made here and/or shown on Ren'py. Just looking at the Games Page you can see how each game has there own visual flare and look. Just from the screen shots each game stands out on its own. Not one overlaps or looks the same, even the three black and white games are distintive from each other.

Even with lots of options and line width change, I can't help fear that each game would lose its own individualism if the Super-Paperdoll system was used too much. I'm just not fond of the idea of a future with the game page with a bulk of screen shots with a kinda uniform look to them. Thats just my feelings on it though.

Again though I can defiantly see why writers would love such a tool at there disposal.

---

Actually making the thing would defiantly be difficult and a hell of a lot of work. I've seen flash programs and character makers, as shown by Monele (And oooh I love the art in Tokimeki Memorial Online - I'm currently watching the anime series. Pee Pee Pyro Pee!), but I couldn't imagine all the work needed to make such a program. My guess is it'll be an indivdual program you run, make the character, choose the expressions and it makes/saves the png files for you to import?

[DaFool]

Just a question : *who* is going to do all this? ô_o... That's an incredible amount of work.

Uh, me? I'm no ace artist, but all the more I need this tool for myself at least. My batting average for drawings I consider 'good' is pretty low, and I'm ambitious, so...

without having to wait for those damn lazy artists (;_; I'm sorry) and if you manage to pull it off more power to you, but from my perpective I actually hope it doesn't get used too much when making visual novels around here.

To explain:
For me a lot of the allure for visual novels is the artwork in it. I know everyone has there own personal peference on how much its the story and how much its the art, but for me I do tend to look a lot into the graphics. This is even true for the games made here and/or shown on Ren'py. Just looking at the Games Page you can see how each game has there own visual flare and look. Just from the screen shots each game stands out on its own. Not one overlaps or looks the same, even the three black and white games are distintive from each other.

Even with lots of options and line width change, I can't help fear that each game would lose its own individualism if the Super-Paperdoll system was used too much. I'm just not fond of the idea of a future with the game page with a bulk of screen shots with a kinda uniform look to them. Thats just my feelings on it though.

Again though I can defiantly see why writers would love such a tool at there disposal.

I absolutely agree that custom individualized drawings for each game is the way to go. In that sense the OEL gamemakers already beat the translated Japanese al|together games, since we've seen enough to see which are reused sprites / backgrounds.

However such a tool may be a godsend to those who really have a story to tell, but may worry that no one would want to read if theres no halfway decent graphics to attract people in the first place. They may get discouraged if they find minimal interest by artists to work on their project, so where do they turn to? Sure there are quite some here who have deviantart accounts, but the majority of deviantart people don't really do Visual Novels. I've managed to come across near-complete games in terms of script, but which were never completed since there were no artists interested, which I think is a sad thing.

The way I aim to reduce graphical sameness is to make things look similar but not exactly the same. Just like you look at a Navel game (Shuffle!, Tick Tack, etc.) and you come across the same character designs in other games -- sure you know they all look the same and the only thing different is their hair, but they're different games entirely and if you knew Japanese you may be willing to try them out each for their own merits. The key is that there's a halfway decent style to attract you even if it looks awfully familiar.

My guess is it'll be an indivdual program you run, make the character, choose the expressions and it makes/saves the png files for you to import?

The Alice program by PyTom is already made and is pretty sufficient for this purposes. I just aim to put more variability through lots of options. If I do ever need further help, there's plenty of programmers here, I just need to show the drawings, so

[dizzcity]

Speaking of which... the thing I find most disadvantageous about using Alice and Betty is that there is no Alex and Benny to go along with them. I can't tell you how many times I've wanted to include a MALE character, but simply couldn't find one to fit. So do please please please make sure both sexes are represented.

-Dizzy-