Website powered by

Mara in color (with AI)

Ever wondered how I really imagine Mara's colors? Well, here's a little preview of it (although the final game will come to be in that distinct green GB color).

Play my Demo of "Rosa Cupiditatis" at "itch.io":
https://nova-odos.itch.io/rosa-cupiditatis-demo-kickstarter

Possible Q - A
1.- Why did you use AI to make an art of your own character?

A.- Because AI is a "thing" right now and I need to update my workflow to understand how this technology works. Also, it is kind of fun fixing AI's gibberish.

2.- What AI platform do you use to make AI images?

A.- PixAI

3.- Why not Midjourney or any other platform?

A.- Because I'm not paying for generating images I could do myself, PixAI is very anime/manga oriented and I like that stuff and although PixAI has some restrictions I've become used to the rules and usage of this platform.

4.- Aren't you cheating by using AI to generate this sort of content?

A.- It took me around two to three hours in doing the whole editing (there really was a lot of fixing like the lips and hands for example) and many art pieces are done in a "collage" way just like this one.
The original input image was done by me.
This work is done mainly for fun/learning purposes, I do not intend to use nor rely on this workflow to achieve the final art for my game.
And also, AI for making a Gameboy image background with a 192 tile count limit is way beyond AI's capability as of today.

5.- Why haven't you posted anything of the game's WIP?

A.- I'm working hard on battle mechanics. I'm testing my ideas quite a lot and I don't want to post an empty update of a thing which might change in one week. But make no mistake, I'm actually pretty hyped to post the first three battles.

6.- Aren't you afraid people might actually like the AI art better than your own art?

A.- Nah, nowdays who cares which drawing is better. I draw because I like it, not because I want to be better than AI.
Besides, I'm more worried about keeping all my variables in check for my game than some dumb deep learning algorithm designed by another human, who maybe (just maybe) doesn't know how to draw ;)

Seriously...I need to setup a special LoRA to just get "good hands", gimme a break.

Final image AI + Digital art

Final image AI + Digital art

Step1: Original art is 160 x 144 pixels. So it was upscaled in Affinity Photo.

Step1: Original art is 160 x 144 pixels. So it was upscaled in Affinity Photo.

Step2: Image generated with AI based on "Original Input". In this image I only liked the hands and arms.

Step2: Image generated with AI based on "Original Input". In this image I only liked the hands and arms.

Step 3: Image generated with AI based on "Original Input". In this image I liked the clothes and the face.

Step 3: Image generated with AI based on "Original Input". In this image I liked the clothes and the face.

Step 4: I edited the prior two AI images and merged them into one single composition and fixed many issues I didn't liked from the very beginning.

Step 4: I edited the prior two AI images and merged them into one single composition and fixed many issues I didn't liked from the very beginning.

Step 5: I took a shot and fed the prior image to the AI and got this new output.

Step 5: I took a shot and fed the prior image to the AI and got this new output.

Step 6: Edited in AffinityPhoto/Designer the AI generated image and got a more interesting result.

Step 6: Edited in AffinityPhoto/Designer the AI generated image and got a more interesting result.

General workflow idea.

General workflow idea.

Liked this work?
Check out this character in my game demo "Rosa Cupiditatis" at "itch.io".

Liked this work?
Check out this character in my game demo "Rosa Cupiditatis" at "itch.io".