Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines

Advice for Using AI artworks in Visual Novels

A topic by NOIR created Oct 23, 2022 Views: 1,194 Replies: 1
Viewing posts 1 to 2
(1 edit) (+2)

Hello all, 

Over the last week or so, I've had a bit of a revolution in terms of my game development thanks to a friend introducing me to Novel AI. I want to be clear that although I'm going to say some glowing things about Novel AI, I have no affiliation with the company, etc, and I'm sure there will be competing AI's out there that can get the same results. 

I am somebody with very little artistic talent that nevertheless decided to make a visual novel in the hopes that one day I would be able to work with artists who could replace my (terrible) placeholder artworks. The last week of experimenting has taught me that I probably no longer need a human artist if I continue making my relatively simple game. 



So, the immediately obvious strengths of using AI artwork:

- The speed - It's practically instant to generate artwork.

- The quality - Many of the artworks created are extremely visually impressive

-The imagination - This is an aspect I hadn't considered when starting, but the AI can bounce back ideas that you hadn't previously considered and expand your ideas in interesting directions by its own initiative. Sometimes, it does this by mistaking your commands, but this can be an interesting exploration of possibilities.

However, there are also clear limitations:

- The rigidity - Of course, in visual novels you want to be able to show characters going through a range of emotions and poses. In my experience, this is one of the large frustrations with AI artwork: It is currently not very good at taking a character and replicating them in slightly different forms. I've found that even getting a picture to raise a hand slightly, is a near-impossible task. 
-You have to develop a skill for 'fishing' with keywords - I've learned that communicating with the AI via tagwords is a skill in its own right. You need to go into it with a serious idea of exactly what you're looking for and how you'll be manipulating the image into fitting with your game. For example, I would highly recommend including the tags 'Simple background', 'green background' into every single search you make in other to make it easier to cut out your character sprites manually for use in the game as transparent PNG's. 

- Fitting things together is a nightmare because styles clash - For this, the only workaround I've found it to incorporate AI generated artworks into Blender, using the AI-generated work as a plane with an emission shader that casts a light onto other objects within the scene. This may sound difficult, but I think it helps brings believability to the scene. 



Finally, I'd say that the biggest limitation is the LONELINESS of not having another human to bounce ideas off of... I know this is a strange way to end things but in my search for artists, the thing that had been exciting me all along was the spirit of collaboration and community. I feel like AI is incredibly powerful and has been a huge help to me but I also feel guilty for all of the wonderful artists out there who may feel neglected and threatened because of it. 

(+1)

Also, HANDS. The AI is terrible at HANDS and FINGERS. Everything it comes out with looks like it's from the Manga Parasite. 

For most issues, I feel you can 'fish' for solutions and just keep retrying until you get a good result. Not hands though! I just draw them separately on Medibang Paint and hope that nobody notices.