Ai-assisted anime-style monster vore.

[taylorsm]

Been experimenting with NovelAI and posting these in some general ai-art thread. Decided to compile them all here. Will update with new contents when available. Mostly going to be anime girls getting gobbled by monsters + internal shots.

I feel like I'm starting to get the hang of this. If you look at the progression, I think there's gradual improvement.

[Hereforvore]

I know plenty of people are concerned about AI art for a multitude of reasons, but I can't helped but be impressed that this can be done. I really like these

[Undue]

These are pretty awesome.

[VMachine3000]

absolutely phenomenal, i've been on my own journey with NovelAi, any tips you care to share? your internals are just so darn crisp

[taylorsm]

absolutely phenomenal, i've been on my own journey with NovelAi, any tips you care to share? your internals are just so darn crisp

Heya. Glad you like those pics. I can share some tips that worked well for me:

1) The AI doesn't understand vore or typical vore tags, so use tags it does understand. For example, being trained on Japanese style art, it is really good at tentacles (lol). If you want organic looking objects and backgrounds, you can try tags like (detailed tentacles background) or (girl wrapped in raw meat tunnel). Another example is that the AI doesn't understand digestion, but you can use (sweat) or (wet) to splatter fluids everywhere. (Wet) in particular tends to make things partially submerged, which you can then pretend are stomach acids.

2) Use inpainting. Unless you get super lucky with a prompt, the more complex scenes need to be built piece by piece. For example when I wanted to draw that girl in the monster's mouth, I inpainted the maw and teeth until I got a good one, then inpainted the girl until I liked her pose, then inpainted finer details like hands and eyes. The AI is really bad at hands in particular. Sometimes I need to inpaint them dozens or even hundreds of times before I'm satisfied.

3) Sometimes you need to do a bit of manual drawing to push the AI in the direction you want. The good news is that the AI can accept MS-paint tier art as input, so you don't need to put in much effort. For example, to get started on that last pic I started by prompting the girl and then scribbled the monster around her:

Good luck with your AI wrangling.

[taylorsm]

Was attempting another maw shot but with the tongue licking all over the prey. Unfortunately, between the weird angles and complex interactions, I couldn't get a coherent result. Posting a low-effort stomach scene for tonight instead.

[Altimos]

Was attempting another maw shot but with the tongue licking all over the prey. Unfortunately, between the weird angles and complex interactions, I couldn't get a coherent result. Posting a low-effort stomach scene for tonight instead.

I would be interested in the tags used for this...
As this is a great setup for stomachs, I'd like to see if I can run the same tags along with fluid mechanics

[taylorsm]

Was attempting another maw shot but with the tongue licking all over the prey. Unfortunately, between the weird angles and complex interactions, I couldn't get a coherent result. Posting a low-effort stomach scene for tonight instead.

I would be interested in the tags used for this...
As this is a great setup for stomachs, I'd like to see if I can run the same tags along with fluid mechanics

(((masterpiece))),best quality, ((illustration)), nsfw, (pose), full_body, (detailed face), {messy hair}, 1girl, {wet}, fetal_position, (beautiful detailed girl), {beautiful detailed glowing eyes}, {wet nude}, (nude girl wrapped tight in detailed raw meat hole tunnel), (dark red detailed raw meat background), {{medium_breasts}}, solo,wet, sweat,cinematic_lighting

However, I had to guide the AI a bit by drawing an outline of the stomach manually and passing it through img2img a couple times.

[Altimos]

Was attempting another maw shot but with the tongue licking all over the prey. Unfortunately, between the weird angles and complex interactions, I couldn't get a coherent result. Posting a low-effort stomach scene for tonight instead.

I would be interested in the tags used for this...
As this is a great setup for stomachs, I'd like to see if I can run the same tags along with fluid mechanics

(((masterpiece))),best quality, ((illustration)), nsfw, (pose), full_body, (detailed face), {messy hair}, 1girl, {wet}, fetal_position, (beautiful detailed girl), {beautiful detailed glowing eyes}, {wet nude}, (nude girl wrapped tight in detailed raw meat hole tunnel), (dark red detailed raw meat background), {{medium_breasts}}, solo,wet, sweat,cinematic_lighting

However, I had to guide the AI a bit by drawing an outline of the stomach manually and passing it through img2img a couple times.

Thanks for the tags and that last tidbit... I've been trying to work with img2img, trial and error for me.

[taylorsm]

Experiment with layering 2 images in post-processing for x-ray view.

[taylorsm]

As gif.

[VMachine3000]

Experiment with layering 2 images in post-processing for x-ray view.

I really think that post processing and indrawing is where this stuff really shines, really like what you've been able to accomplish

[Randomdude5]

This is really good. I have never messed around with AI, but I wish you all luck, and am eager to see what other cool stuff can be created with it.

[taylorsm]

Building of the previous idea, did a x-ray spot of a mid-swallowing scene. The snake is kind of goofy, but the AI is pretty bad at snakes, so I don't think I can make it any better.

This piece was really complicated because there was 3 main parts: "snake", "girl internal", and "girl external". It was really annoying to get them to fit together smoothly. I probably won't be attempting anything like this again unless I find myself with a ton of free time suddenly.

EDIT: Shrunk the snake's head a bit and I think it looks a lot better.
EDIT2: Adjusted girl's position slightly.

[taylorsm]

This image is similar to some earlier ones and was generated in the same batch. However, I noticed that it had some really nice texturing and details. There's even a little channel of fluid at the bottom.

[HanabiV]

Experiment with layering 2 images in post-processing for x-ray view.

How do you go about creating the girl part of these images. Do you just regenerate until you get a pose you like, or do you start with an image?

[taylorsm]

Experiment with layering 2 images in post-processing for x-ray view.

How do you go about creating the girl part of these images. Do you just regenerate until you get a pose you like, or do you start with an image?

I generate girls until I get a useable pose. Then I paint a blob over her and try to get the AI to turn it into a pred.

[taylorsm]

More layering. This time I tried to generate the pred first, then fit the prey in afterwards. This means I can generate a nicer background for the externals.

[taylorsm]

[taylorsm]