Testing ChatGPT's Image Gen 2.5: a subtle half-step upgrade
Aizkmusic · x · 2026-09-09
Testing the new image gen 2.5 with a fighting-game character select scene and the prompt "make me laugh," the author finds improvements that are slight and subtle—hence the half-step version bump. Results still clearly look AI-generated, but "it's the worst it will ever be."
Related event: GPT-Image-2.5 quietly rolls out in ChatGPT with modest gains, users report(6 posts)→
More from Multimodal
- NoSpoon creator shares final musical preview of The Necromancer's Apprentice — Kyrannio · 2026-09-09
- First full short made with Seedance 2.5 shows major leap in AI video realism — DavidmComfort · 2026-09-09
- Midjourney v8.2 sample image: Slots-themed generation showcase — chrisfirst · 2026-09-09
- GPT-6 Astra reportedly built a full 3D human cell model in 30 minutes — Eric520CC · 2026-09-09
- Zero-editing music video: local MiniMax song plus GPT-directed MV pipeline — Successful-Roof-9220 · 2026-09-09
- World Labs demos real-time next-view prediction with accurate 3D consistency — drfeifei · 2026-09-09