If you come up with a story and send me 1-3 good pictures of your character, I can make a comic with your character too

Basically, I don’t really have any ideas that I wouldn’t be able to do - it mostly just depends on how complex the work is. Some things can be done relatively quickly, especially if there’s only one OC. Some things take much longer. For example, if we have a scene like “the character is sitting and looking at a computer screen with something shown on it”, then the image on the screen would just be a regular inserted picture - I wouldn’t draw a full program interface from scratch, since that takes too much time

So overall, this is interesting to me. Maybe I do have some limitations, I just haven’t run into them yet~

Simple scenes like “sitting on a chair”, “drinking tea”, “walking down the street”, as well as gestures and facial expressions - that’s totally my thing

@CloudWings
I more question how you could create four versions of a character that didn’t exist before. What I mean is, that is a custom OC, not a show character. The art style looks consistent, the colors are consistent, and the various features of the character look consistent. I would think that would be difficult for a LLM to do.

I don’t know how I can include generated images here as examples, but I’ll try to explain the process

First, I write the story in text form - what exactly should be shown in the comic. In other words, I come up with the idea for the piece in advance, and I also write all the text beforehand. This helps me understand the overall composition and flow

Next, I generate several images of the character and choose a version that looks relatively consistent. After that, it becomes more of a technical process - adjusting hands, eyes, and mouth, and refining details

In the end, I’m left with four generated images that I want to include in the final comic. Then I add text and other elements on top. In this case, things like “REC”, camera labels, and similar details are part of a camera interface that I recreated from scratch using a reference. All the text is also added manually, and the final composition is assembled together

That’s how the final image comes together: my idea + the neural network + additional manual edits on top (text, overlays, and other details)

@Eragor14
Some of the lines on the wings look a bit weird for something a human would do but yeah it’s hard to tell. Course any good AI art these days has some amount of manual editing on top of it to clean up any artifacts so if they’re doing that it can make it less obvious.

Also if you crank up the contrast a lot you can see there’s a bit of noise in the otherwise clean background, given this is a PNG there shouldn’t be jpeg artifacts or anything like that which is probably the biggest indicator (since a human artist would generally have a perfectly clean background, but AI struggles to get it exactly perfect). It does have some jpeg-esque artifacts though around the character so who knows though, maybe it was a jpeg originally and re-encoded, but I wouldn’t be surprised the AI just learned to generate those sorts of artifacts itself or something by being trained on a bunch of jpegs.

Is…this really AI?

This doesn’t look like AI in the slightest. It looks like someone drew this.

That makes me very concerned if it is AI, though…