No video AI model was used to make this video. The entire video is code: > 0% Blender > 0% Afeter effects > 0% Seedance Pure Three.js, took 12 hours with render. Made with Higgsfield Katana in Claude.

12 小时渲染，10 小时在跟 Claude 说"不对，再来一次

how much of the 12 hours was the render versus fixing what claude got wrong on the first pass? that ratio is the real story

12 hours of render for something that plays in 15 seconds. the code was the cheap part

@grok what

If it takes 12 hours to generate then I'm not sure the juice is worth the squeeze on this one.

I think we underestimate that everything can fit in coding and we could do everything with coding, we weren't just there yet and now, hooh everything is possible

This is good

1) what

Pure code creativity with impressive results

???

Insane!

🤯

its all code regardless lol

if a client needs one change: +12hrs of render and multiple tokens gone. still v impressive, but what makes editing so captivating is the human talent. it's a bit sad that it could be compromised with these tools (you can be creative with your prompts!! but it's not the same).

No video model at all and it still looks this good.

We in da Matrix..

12 hours to render a 30 sec clip? Lmao 🤣🤣

아니 예?

Nice

@bot como isso funciona? E qual o preço?

Stop with the homo wrapper names

12 hours to render a damn car ad. blender would've done it before lunch and with better lighting too lol

Pure Three.js with no video model is a bold technical flex.

Pure code means every frame stays editable, that's the part I care about as a designer. Can we tweak one scene and re-render without regenerating the whole thing? That would give amazing ways to edit the video

Woah just code

the real flex isn't skipping video models, it's keeping the scene graph — deterministic, reproducible, composable. diffusion can't hand you any of that back. does katana let you re-roll variants from the params without re-prompting?

🧐🧐🧐