ComfyUI fails to generate Megumi Fushiguro; ChatGPT walks through prompt fixes
Potential_Save · reddit · 2026-11-22
A ComfyUI newbie asked ChatGPT to guide him through generating Megumi Fushiguro at the beach holding a coconut — and got a cast of distant relatives instead, plus a 30-min-per-image Qwen Image 2.1 two-character scene. The screenshots show ChatGPT's debugging: fixing the diagonal camera angle, then diagnosing age, proportions, and the coconut-turned-green-cup, with concrete fixes (keep Megumi LoRA at 0.80, revise prompt). Funny fail with a useful troubleshooting workflow.
More from Fun
- AI researcher's 'billion dollars' tweet was actually her cat walking on the keyboard — mmitchell_ai · 2026-11-22
- GNU-style joke: 'What you call wandb is actually Coreweave Forge' after acquisition — charles_irl · 2026-11-22
- Claude's refreshingly honest answer: blood donation's health benefits are overstated — OnlineInference · 2026-11-22
- Waymo rider: I feel relief crossing in front of a robotaxi, not a human driver — Afinetheorem · 2026-11-22
- Matt Turck mocks the trend of tweeting excitedly about AI and the Riemann Hypothesis — mattturck · 2026-11-22
- Brautigan's 'machines of loving grace' poem recirculates in AI circles — ZeroStateReflex · 2026-11-22