underwood(@underwoodxie96):I created this video using GPT Image 2.0 and Kling 3.0. I also tested Seedance 2.0 and Google Omni. They're much faster for generating transformation videos, but in my tests there were often one or two shots that felt slightly unnatural or had that unmistakable AI-generated look. For this project, I wanted the transformation to feel as smooth, natural, and cinematic as possible. So instead of generating the entire sequence in one go, I used a start-frame / end-frame workflow, created multiple short clips, and refined the pacing and transitions in CapCut. It took more time, but the final transformation feels much more seamless, believable, and film-like. Full workflow + prompts:

2026.06.01 11:48

I created this video using GPT Image 2.0 and Kling 3.0. I also tested Seedance 2.0 and Google Omni. They're much faster for generating transformation videos, but in my tests there were often one or two shots that felt slightly unnatural or had that unmistakable AI-generated look. For this project, I wanted the transformation to feel as smooth, natural, and cinematic as possible. So instead of generating the entire sequence in one go, I used a start-frame / end-frame workflow, created multiple short clips, and refined the pacing and transitions in CapCut. It took more time, but the final transformation feels much more seamless, believable, and film-like. Full workflow + prompts: