It's been a few days so it's time for an update.
Basically, I've
wasted spent 10-15 hours over the past few days on image generation and editing, with a three different goals:
First goal is to get a set of basic "tachi-e" (standing sprite) images of the main characters that I can use as a basis for making VN dialog scenes.
Working with AI generated images is very much a two edged sword - the good part is that I can at least create *something* that could be used without forcing everyone to scratch their eyes out in horror. The bad part is that *everything* needs tweaking, and the closer you look the more imperfections you see. Minor clothing mutations like missing zipper pulls or collar lines, noticeably different hair styles, lopsided boobs, varying color palettes. I struggle with perfectionism to start with so it's hard to know when to stop grinding on a single image and move on to the next. However, their is nothing for it but to just keep working at it - no magic wand* exists, and either I do it myself or I have to pay someone else.
I thought I was reasonably proficient at image editing, but I've been learning that there is a whole 'nother toolset which I need to master. There's only so far I can get with my old faithful PaintDotNet, lol. Photopea has been what I am trying to learn for more effective tooling so I can more effectively cut-n-paste of elements from multiple images into something that is not horrible, but there is a hidden world of "photoshop" work-alike shortcuts and workflow that is not easily discoverable. Lots of tutorial watching already and more in my future. Smart selection, feathered cuts, layers, color palette match, color picker+small brush to fix seams and weird AI stuff... the list of required skills grows daily.
Here's a couple of samples I picked out - and I must hastily clarify that there are pre-edit, thus the variations in her jumpsuit uniform design.
Second goal is to produce some location background images. This has been far harder than I would have wished - I guess the LORA that I am using is not trained very much on background scenery, and especially not "starship command deck" images.
I did manage to get this guy though...
Third goal is to get a replacement set of images for the heroine when she is in the 'Hole. This is to use in developing the UX feedback when playing your action cards - I don't want to publicly paste gameplay samples based on my pre-alpha version where all the art has been "donated" from another game. I also wanted to lean into the multi-image animation sequences a little harder than the original - a 4 frame image loop is better than 2, and 8 frames is bettter than 4. I'd actually hoped there had been more progress around this aspect in the AI image gen world, but so far I've not found good examples - all the "animated gif" stuff is weird psycho-dream morphing horrorshow junk.
For the glory hole asset work, for now I've got an invited guest model helping me out - not aiming for final assets production at this stage because I'm still learning the tools and workflow. Here's a sample image...
All this time spent on artwork has lead to the code dev being on hiatus for the past week or so, according to my commit log. It's not wasted time, however, because a game with just code and text is not what I want to create - a certain minimum level of artwork is a prerequisite.
Anyway, that's all for now. I've just ticked past the 200 hours mark in my work log, counting research, shit posting in here and other threads, coding, writing concepts and story dialog, and artwork. Only another few thousand hours to go!