Image gen censorship is inadequate!
retained text
I'm developing an RTS game in urban combat setting. I use Claude code and the only reason I still keep my GPT plus subscription is image generation. But when I needed to create unit portraits it became totally unusable due to its paranoid violence filter. I made the attached picture, then I asked it to generate similar characters, preparing to throw a stone, holding rebar in a battle pose, with hand-made crossbow, etc. And each time it refuses due to violence filter! It admits itself, that this is false alarm, but can do nothing. Suggests to change the style into less photorealistic. But why? There is no blood, no killing, just men with weapons, why can't I have them? I have to use my free gemini account and it has 0 issues with that. I'm not just satisfied with quality. Have never faced such problem before. Don't actually know what to do.
document fetched 2026-11-10 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xrkqo3/image_gen_censorship_is_inadequate