Users report GPT gradually corrupting detailed image prompts across turns

apostrophefee · reddit · 2026-09-25

A user complains that GPT gradually rewrites their detailed H3 image prompt across a conversation, adding and corrupting details along the way. They need an LLM to handle long, precise prompts and find the experience frustrating — contrasting with the popular belief that AI just magically produces whatever you want.

Original post →

More from Fun

Fun channel →