18 / 25
May 2023

If you gave me $500 a month to improve my comic, I would probably spend it on things like getting a new computer, monitor, and tablet. And use the rest for just making sure I have good food to eat and I can stay healthy. My biggest struggle with working on my comic is me feeling sick a lot.

I don’t know if I would spend the money on fancy programs. I have used TinkerCAD to make 3D objects, which is free. I have seen people use the Sims for planning out houses. And other people have used Blender. I have also seen great designs come out of SketchUp.

yeah I was gonna say, I wouldn't offload any creative or highskilled tasks to AI. That's the fun and challenging part of making comic. Like, it's being able to create my own world with my own art is part of why I do it? But there a lot of brainless labor that AI might be able help with, like flatting or line art and colors clean up-- whether I use it depend on quality because for 500, I can hire a human to do it.

I guess if I were to delegate from that list, maybe if an AI can produce a basic and editable premade 3D scene for me as a starting point? That sounds unrealistic though.

It seems like there are already some things out there to create 3D worlds though. A guy sent me this one recently, check this out:

You type in a prompt, draw an outline and it generates this landscape around you. It feels very "game-like" but I don't think we're too far away from having something like this for normal 3D scenes which means it'll also make its way into the webtoon industry

im pretty sure thats AI generating skyboxes-- image projected into a box to create a 3D illusion-- rather than actual 3D models. It's v. useful for videogame design, but not as useful for creating comic because comic artists need actual 3D model to replicate multiple perspectives of one setting.

I also don't know where they're getting their data to create those skyboxes.

By realistic, I mean I can't see how AI learning could help us generate 3D scenes right now. Scene generator is already thing in 3D programs. Video games company uses it to generate 3D maps-- I.e Minecraft seeds. The thing is, programers hand create and categorize all of the assets use within a scene (chairs, tables, etc). When I make a 3D scene, I search for, and import these assets into the scene and set them up where I want them.

AI could support this scene generation process either by learning how to actually generate a 3D asset (I ask for a 3D model of a table and it delivers), or it can help categorize the assets to help generate scenes accurately (it knows that a "classroom scene" needs to have tables and accurately generate the table on the ground instead of ceiling). I dont think AI is anywhere near this level yet, and for it to learn any of this still requires a massive database which might have to be stolen.

Wait, but isn't there already like a million tools out there that can do flat coloring or at least make it a 30sec task? For example things like (https://petalica.com/index_en.html6) or just using the bucket tool in Photoshop come to mind for me or some extensions from Clip Studio.

Is there something these options aren't doing well that makes you feel this way?

With the ones that exist, like the linked one, I'd still need to click all the areas I want certain colours with the selected colours, just like I do with the fill tool then manually tidy up the bits that aren't to my liking. It'd save me zero time unless I spent a huge amount of time and effort training a bespoke AI to recognise my characters and apply the appropriate flat colours based on which character it was.

Yes, the tech might seem very clever when it's colouring in one standalone illo, where it doesn't matter about applying the same colours consistency across multiple drawings of the same character, or maintaining or deliberately using light direction in a scene... but it's just not useful in a comic, for all the same reasons your background-drawing AI tool isn't helpful; it's made without considering that one of the reasons AI can't do comics is that it treats every panel like a new image, because it's applying the algorithm afresh to a new problem. The consistency is always the hard part.

Do you have an example of a line-art vs. flat colored image? I'm trying to imagine it but Google isn't doing me any justice to picture what you're talking about (@Lensing, @darthmongoose)

(since this question is geared towards backgrounds)

For me I'd be allocating the last step: editing images to fit my anime/manga style.

I use Sketchup/Clip Studio so I don't have to redraw the backgrounds, but I do need to manually add effects/retouch some objects to my liking. That's the biggest timesinker for me, when it comes to backgrounds.

My current process is:
1. Export background scenes from SketchUp, using a SketchUp plug-in (the lines, textures, base colors are all its own separate layers)
2. Paste them to my panels in Clip Studio
3. Adjust the overall colors by applying a color filter, depending on the mood.
4. Add simple shadows, if any
5. Enhance the perspective via the fading white airbrushes
6. Make the line layer lighter so it doesn't stand out too much
7. Apply any other retouches/effects (such as the plant leaves, glass shine, etc.)

Sometimes I skip a step or two depending on how much time I have.

The pre-made 3Ds are needed especially for someone like me, who doesn't care about creating unique backgrounds lol.

But I don't buy them too often to need a set budget each month and/or need someone to do it for me. Either I make do with what is available for purchase, or scour the internet to put together a model myself. Though it would be nice if I could type in keywords like "middle class American apartment" or "fantasy city golden gates" and get a model that way.

I don’t see this being too helpful for comics due to the lack of control.

Backgrounds can be as unique and iconic as the characters who inhabit them. It’s something that can’t always be produced by just being random.

Maybe this feature might be more useful for people making Visual Novels.

Or it might be like a lot of tech, it looks cool but has little function outside of that. My brother used to make programs like this, they would do interesting things but had no practical function.

I wouldn't let anybody do my backgrounds for me. It would be as unthinkable as letting somebody else drive my car.

But seriously. I draw because I enjoy drawing. I do everything freehand and as I get more acquainted with my art programs I improve. Letting somebody else do the work would take the joy away.

As far as a $500/month budget, I'd buy one of those fancy 4k 21" drawing monitors and make the payments on it with that $500/month :yum:

I would spend 500€ to get advices from people who studied manga narrative: there are few manuals who talk about baloon placing and how to focus on the general flow of the page. This would help me a lot, I struggle when I create the storyboards beacuse I'm not sure where or how place the elements to give that particular emotion.

For backgrounds and 3d models I use Sims 4 and look on Google maps for the references.

Oh, what a great reference.

I really loved that show growing up. In France that was one of the only US cartoons we got alongside Futurama and every day during our lunch breaks me and my sister would come home and eat while watching The Simpsons. Really brought up great memories and I totally get what you're saying. This kitchen is iconic.

I agree with you on the tech front. Many things are "cool for 10min" but aren't useful beyond that point. That's definitively the fate I'm trying to avoid because to be honest I'm pretty addicted to webtoon & manhwa and I've noticed that artists will often struggle to post every week; unless they take significant tolls on their health.

In reality, I'm selfishly trying to make tools to help them make more chapters so that I have more to read :joy: This is not a selfless quest hahaha

If you can't even envisage the problem, I'm struggling to understand why you're trying to make an app to fix it. I feel like you're approaching the development of this whole system backwards, by deciding you want to make a solution for something that people will buy, and then hunting for a problem that solution will potentially apply to...

But okay, this is what one of my pages looks like before I colour it:

And this is what it looks like after:

I use a rough palette like this to keep track of which characters have what colours on them, and update this every time the characters get new outfits (though one-off outfits in flashbacks or on minor characters I just deal with as they come)

So... the reason I can't just get a computer to do it, is because the computer does not know who Rekki Lune is. The computer might recognise that there are drawings of people here, and might apply skintones to them when it detects a face, for example, based on machine learning, but it won't know to spot my characters specifically and apply the right colous to the right areas so that each panel has those characers coloured consistently; ie, so Rekki's pants are red, her skintone is that specific colour etc.
Since most of these areas are flat fills, a large part of the process is just clickng an area to fill it...meaning that a tool where I select a colour and click that area to "auto colour" it saves me no time. It might save time if I was doing some sort of soft-shaded style, or if I was making a comic where for stylistic reasons, consistency wasn't an issue (like say if it was in the style of "V For Vendetta"), but for me, it won't. I'd need to keep correcting its choices about colours, like "no, Rekki's skin needs to be this colour," "no, this part you sensed as being skin is actually tights..." It won't save me time.

If your goal is to make a product to sell to comic artists... which er... may be a bad choice of market, we're not the richest bunch of people.... You really need to sit down with some and do proper UX research; watch streams of webcomickers at work, tutorial videos showing processes, look for pain points.

In verdi's slight defense, it does feel like a lot of comic artists do gripe about backgrounds. It's definitely one of the more common things I see comic artists try to outsource (as in, "hello, I am looking for an assistant who draws backgrounds! Let's 'collaborate'!!! :D"), and AFAIK in the world of successful manga-ka, drawing backgrounds is often something relegated to an assistant. I think there actually MIGHT be use for a browser-based tool like Clip Studio's "extract lineart" feature. Like, something that would allow you to upload your storyboard, overlay a 3D setting on it (assembled from available assets - ideally you'd have an indie tier where people can use the assets for free, and a paid tier for people who make comics for profits), then extract transparent lineart and stick it in your panel. I know Clip Studio has that functionality, but it would be cool if it was more accessible/they had some competition.

I personally wouldn't use this tool because my art is heavily stylized, and I need the background to be its own, matching character. But I can definitely imagine people whose stories are more heavily driven by character interaction, where setting is secondary, finding something like this useful (In fact, there is a story I'm not drawing just because I can't be bothered with the background, so... this may actually be directly useful to me)

I would also definitely be interested in something that could flat pages for me (it's actually a big pain point for me, as a disabled artist with drawing time limited by my injury), but like darthmongoose, I do not believe the current approach to generative tech is in any way capable of something like this. I've tried getting it to fill a couple of my pages and the results were decidedly nightmarish. ETA: lol3

I already use that. It helps somewhat, but can only really save me so much time. :sweat_02:

I wanna show you what I mean, since I just finished working on exactly this today.

Here's a 3D Sketchup model that I made. Unlike the skybox, it's an actual 3D object that allows me to rotate at any angles I want

Which I use as the foundation for my my backgrounds.

A model takes me ~30-60 minutes to build on a free software. It's quick and rough because mainly use it to get the perspectives and foundation of the scene; I can add details on my own. Some people take much longer because they add all thee details they need to their model. Some people import fully rendered scenes that others have made to speed up the process.

For step 2 of your suggestion: Im not sure how AI would help find premade 3D scenes. 3D assets are categorized by their creators. Sketchup has the "Warehouse1" for example, where you can just search for premade asset. I also don't think AI is at a point where they can recognize 3d objects, let alone full scenes?

For step 3: I don't think AI can generate 3D models right now either. To do so would require a massive database of 3D assets. Unlike 2D images, which can be swipe on google, 3D assets aren't openly available. It would require partnering with company like Sketchup to allow you to take premade 3D assets and put them into a database. And correct me if I'm wrong but AI can't process 3D models. The way Art AI work in very simple term is is by processing images as code, scrambling said code, and produce a new code that results in an new image. 3D assets are vastly more complicated than images.

Step 1 and 4 is simple. Storyboarding takes a few minutes a panel but also require knowledge of composition in relation to your story's narrative (i.e which camera angle will make a character seem more powerful vs weak). AI could help with generating ideas perhaps, but it seems unnecessary. Getting proper camera angles also doesn't take much work. It's just spinning things to the way you want it.

Step 5 is probably the most useful and possible, as others have point out. As you can see. the the screencap of my 3D model is very bare. II would be helpful if Ai programs can render details (i.e line works, wall texture) using my screenshot of the sketchup. BUT that's basically what AI already do and it require data, which, as of now, are stolen.

Thank you for sharing those images! It's extremely helpful and...

Surprise! Welcome to my UX research project hahaha :joy:
1

Nah, but all jokes aside that's one of main reasons why I try to talk the artists; so I can understand more about their process. So far I've interviewed 32 with whom I talked 30min-1hour on the phone to understand their pain points.

I am 80% confident this is right problem to solve. The how I'm still figuring out: by throwing out various concepts and letting them be picked apart by the community until there's one that makes people say "That's it! This is what I've been needing in my life"

If your goal is to make a product to sell to comic artists... which er... may be a bad choice of market, we're not the richest bunch of people

Why make product to sell to artist who aren't big ballers? Well I've already done the opposite and I've made a lot of money but beyond a payout, it makes no difference in my life if some Eng Organization out there is 33% more efficient TBH.

However, if webtoon artists are 25% more efficient, that directly affects my quality of life in a very tangible way :slight_smile: All else being equal, I'd rather make something for an industry I care about and this just happens to be one of them.

Thanks for sharing! I think I have good grasp now and what would be useful to artists now.

I don't think AI can generate 3D models right now either.

It can, it's just that no one has done it. We're not too far away from this reality but like you said, it would require some sort partnership with a large company of 3D assets like SketchUp or Acorn to have access to large amounts of data.

I believe this is ultimately the future of 3D scenes. In 5 years people won't buy one-off assets. They'll have a subscription like with Spotify and the creators will get royalty every time someone uses their asset to generate a scene.

Until then, I will need to work really hard to figure out how to answer to this:

It would be helpful if Ai programs can render details (i.e line works, wall texture) using my screenshot of the sketchup.

In a ethically-safe way.

(Just answering for funzies)

If I were aiming for a polished product, I'll probably spend that 500$/month on hiring someone to clean up my lineart/colouring/etc. But with my current process, I'm honestly not sure what I'd want to outsource. My comic is already sort of a 'minimum viable product' XD

(I guess I would actually outsource backgrounds for bitwam if it were the only comic I'll ever make. The backgrounds (mostly) aren't that important to the story, so I'm cool with (most of) them being generic. But I am planning a future comic where the scenery is important (all the time), and I want to get some practice in :P)