3D Websites Just Became FREE (One Prompt)

5 Просмотры

Спасибо! Поделитесь с друзьями!

Вам не понравилось видео. Спасибо за то что поделились своим мнением!

Издатель Aug 20, 2026

Master Agentic Workflows: https://www.skool.com/earlyaidopters/about
Get the care package FREE: https://markkashef.gumroad.com/l/local-ai-care-package
Work With Us: https://www.promptadvisers.com

This entire cinematic 3D website, the backdrop, the sound, the immersive scroll, was 100% generated with local models on my computer. No APIs or MCPs were hurt in the making of it. A few months ago you needed closed source models for the video, the images, and the website itself. Not anymore.

In this video I break down the three pieces you need to replicate my setup, a local harness (Pi.dev), a local video model (MiniMax H3 through ComfyUI), and a text model you trust (Qwen 3.5 122B in LM Studio, though 27B to 40B is plenty and even 15B works). Then I build a brand new website live on camera with one prompt, a fictional private bank called Meridian, and let my local model design, write, and test the whole thing while we talk.

I am also honest about the hardware side, because nobody else is. You will see my pixelated v0 that took four and a half hours, what two days on a Mac Mini buys you, how a free upscaler closes most of the gap, and what the two dollar cloud render looks like next to all of it. Push your hardware as far as it goes, then upgrade the pixels, not the machine.

The care package is the second link above and it is free. The immersive web, editorial web, and effects skills from the video, my tips for designing and optimizing your own harness, and the raw HTML of the websites I showed so you can have your model of choice reverse engineer them.

---

0:00 - the $0 website (and two more)
1:00 - the three pieces you need
1:45 - which model sizes actually work
2:20 - a harness is limbs for the brain in a jar
3:00 - upgrading the stock harness
3:45 - the three skills that do the heavy lifting
4:45 - go deeper (community)
5:45 - realistic hardware expectations
6:30 - the free upscaler trick
7:30 - what $2 of cloud quality looks like
8:20 - LM Studio + Pi.dev, live
9:00 - the one prompt, explained
10:30 - the build runs while we talk
11:30 - the reveal (and the blank page it fixed itself)
12:30 - where local models go next
13:00 - grab the care package

#localai #minimax #qwen #lmstudio #comfyui #opensourceai #aiwebsite #webdesign #3dwebsite #threejs #scrollytelling #claudecode #codex #aivideo #freeai