Reddit user shares six factors for keeping ChatGPT-generated characters consistent across scenes
SiennaVaire · reddit · 2026-09-10
A Reddit user describes experiments using ChatGPT Images to keep a fictional character's identity consistent across photorealistic scenes and outfits.
Key insight: generating one good image is easy — persistent identity across very different environments is the hard part. What matters most so far:
- A fixed facial description
- Reference images
- Controlled camera framing
- Coherent light direction
- Realistic skin texture
- Avoiding overly complex poses
The author is asking the community how far current image models can realistically go with persistent fictional characters.
More from Multimodal
- Open Actor Protocol Proposed to Standardize Siloed AI Filmmaking Workflows — TheChuckTone · 2026-09-10
- Filmmaker uses Computer Use to drive Blender and editing, calling taste the new interface — taherdhanera · 2026-09-10
- Open-Source Ref2V Workflow Auto-Transcribes Media and Formats Prompts for H3 Video Generation — bstr3k · 2026-09-10
- fal launches Multi-angle for H3 Max: any camera angle from one image in under 3 seconds — gorkem · 2026-09-10
- Valeo fits scaling laws for video diffusion using 5,500 hours of driving footage — abursuc · 2026-09-10
- Zero-cost fix for Minimax H3 plastic skin: drop contrast before VAE decode — Cequejedisestvrai · 2026-09-10