Are there viable careers for Generative AI skills?

I've been learning how to use generative AI for a couple months now, primarily using comfy UI to generate images and videos and I've gotten pretty comfortable with it. Initially I started as a way to expand my skillset since I was recently laid off and haven't had much luck landing a new role in an industry I've worked 15+ years in.

I've been wondering if there is a way to make some income off this? I know people are selling adult content on Patreon and DeviantArt but I'm not looking to get into that, and honestly it seems its already extremely oversaturated.

On the one hand it seems there is a lot of potential to replace content such as video ads that typically have expensive production costs with more economic AI options, but on the other hand there seems to be a lot of aversion to AI generated content in general. Some companies that do seem to be using generative AI are using licensed tools that are easy to use so they just do it in-house vs. hiring an experienced 3rd party. Using tools such as Nano-banana also don't require any local setup or expensive computer hardware for these companies.

In other words, being able to set up an AI client locally and using opensource models like Z-Turbo doesn't really have any demand. So I'm wondering if I should continue investing my time to keep learning vs. pursuing something else?

https://redd.it/1pxzny8
@rStableDiffusion

QWEN EDIT 2511 seems to be a downgrade when doing small edits with two images.

Been doing clothes swaps for local shop so I have 2 target models (male and female) and I then use the clothing images from their supplier. I could extract the clothes first but with 2509 it's been working fine keeping them on the source person and prompting to extract the clothes and place them on image 1.

BUT, with 2511, after hours of playing, it will not only transfer the clothes (very well) but also the skin tone of the source model! This means that the outputs end up with darker tanned arms or midrif than the persons original skin!

Never had this isssue with 2509. I've tried adding things like "do not change skin tone" etc but it insists on bring it over with the clothes.

As a test I did an interim edit of converting the original clothing model/person to gray manniquin and guess what, the person ends up with gray skin haha! Again, absolutely fine with 2509.

https://redd.it/1pxzbh2
@rStableDiffusion

People who are using llm to enhance prompt, what is your system prompt?

I mostly interested in a image, will appreciate anyone who willing to share their prompts.

https://redd.it/1pxzyfc
@rStableDiffusion

Semantic Image Disassembler (SID) is a VLM-based tool for prompt extraction, semantic style transfer and re-composing (de-summarization).

https://redd.it/1py4bxf
@rStableDiffusion