[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/lit/ - Literature

Search:


View post   

>> No.21221921 [View]
File: 1.79 MB, 1280x1280, download.png [View same] [iqdb] [saucenao] [google]
21221921

>>21220163
>https://www.royalroad.com/fiction/57747/simulacrum-heavens-key
What do you gents think about using Stable Diffusion for illustrating your novel? I've tried it, and it is quite fun. I'm yet to personally get a single human with non-messed up hands and face, but if prompt something like the following into the HuggingFace model...

> A very large treasure fleet. A highly detailed, 4k, best quality masterpiece illustration by Greg Rutkowski. Trending on ArtStation.
I get really godly 512x512 background illustrations. Them being this small is not that much of a problem as I've later found out you can use Real-ESRGAN (there are binaries on its repo) to upscale them into something decent. I've gone an illustrated the 170k of my novel and it was a great break in the monotony of just writing all the time. I've had to write a bunch of Python scripts to automate extracting images in the order of appearance from the html file during all that as well for rescaling the results of the upresing to 1,280.

After using SD for a while, I am in awe of its capabilities. Yeah, it does fuck up a lot of the time, especially when it comes to humans, but nonetheless its skill is nothing short of divine. It really gives me a window into what actual AI capabilities are going to be once we move beyond the deep learning phase.

Navigation
View posts[+24][+48][+96]