ChatGPT’s New Image Model Is Insane (GIFs + Consistent Characters)

I tested ChatGPT Images 2.6 against 2.5 on hard prompts, edits, SVGs and GIFs. 👉 Try OpenArt: https://tolt.link/samson-sd25 Thanks to OpenArt for sponsoring this video Every test and prompt from this video, free: delightfuldesign.eo.page/images-2-6 OpenAI says Images 2.6 generates images about 50% faster than 2.5. I ran both models through the same fixed prompt suite, then tested character consistency, markup editing, SVG output, GIFs and sprite sheets, cost at every quality setting from low to max, and moderation. The verdict: an incremental upgrade. The new complexity setting (low, medium or high) is the biggest win, but anatomy and proportion mistakes are still there. I am keeping Images 2.5, Images 2.6 and Midjourney side by side. Every result, side by side with 2.5: https://ltx.aifilmmaker.academy/gpt-image-2-6-sunburst/ Videos mentioned: VectorCraft and the free Adobe clones: https://www.youtube.com/watch?v=x2F3fdn82GU The free, unrestricted open-source image model (MOSS-Image): https://www.youtube.com/watch?v=zO0ZPBj0G3I Chapters: 0:00 The best AI image model just got better 0:33 Flare vs Sunburst: what actually changed 4:55 36 shots, one character: the big test 13:48 Animated GIFs from a single prompt 14:50 Real SVGs you can edit 15:41 Draw on it, and it edits exactly there 16:54 4.5x the price: is it worth it? 17:46 What it still refuses to make 27:10 My honest verdict 💼Business inquiries aisamsoninfo@gmail.com 📧Join my newsletter https://delightfuldesign.eo.page/w7tf5