Had deepseek build a basic image generator frontend earlier, but unfortunately I forgot what a pain in the ass Nix is when it comes to ML so instead of testing I've just watched CUDA and torch be built from scratch for the last like 2 hours
Maybe this warrants some experimentation. I really don't mind ComfyUI since it has a lot of very convenient features that I tend to rely on everyday, but I'm somewhat curious to expand my horizon a bit
I've been using ComfUI app mode for a month or two now and it feels quite nice to not have to look at my graphs anymore unless I need to make a substantial change. But now more than ever I wish there was some different, probably more code-based way to specify my workflows
Like, obviously the graphs work fine. But I tend to feel like (now that there is a way to condense the UI to the essentials) I would much prefer a less boilerplate heavy approach. Probably just preference, maybe Nix brain poison
If you've read this far, thank you!
You can read the original release on pixiv here:
https://t.co/ffA3ZTxdzh
Or here, in japanese:
https://t.co/Sur1Rha9vt
Finally, the reader poll for the next chapter: Who should be at the focus of the next chapter?
Part 4 of my uncensored #AtelierRyza hentai manga.
Content of the next chapter will be decided by poll. Want to leave a vote? Check this straw poll:
https://t.co/bJAqo5yTV1
You can also vote on the pixiv posts at the end of this thread.
#AIArt
1/