Skip to content

Comment on How to generate realistic people in Stable Diffusionparent

Comments

If you prompt the model in the way that pushes it towards low quality amateur photography, the results tend to be more realistic and not glossy:

"90s, single use camera, documentary, of anoffice worker in an open plan office, realistic, amateur photo, blurry"

Results: https://imgur.com/a/GJLqYft

Results: https://imgur.com/a/GJLqYft

Corporate accounts payable, Nina speaking. Just a moooment. https://m.youtube.com/watch?v=4s5yHUpumkY

Is he squinting or am I?

only a prompter would think that looks realistic

Oh a "prompter" is it? No, I'm not a prompter, but if you want to interface with a GenAI model, a prompt is kind of the way to do it. No need to be a salty about it.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.