Pixal3D on 6GB VRAM — Better Than Trellis 2! (Free & Local)

This 3D model right here, I generated with PixArt [music] 3D, the new open-source 3D AI generator that delivers pixel-perfect accuracy from 2D images. Everyone says you need [music] 24 GB to run it locally. I will show you how you can run it easily with only 6 gigs of VRAM entirely for free. In this

video, I'm going to show [music] you exactly how to set it up on your own PC so you can generate unlimited [music] 3D models with the best quality. So, let's get into it. So, what is PixArt 3D actually? If you have been following this channel for a while, you might already know the other open-source 3D AI

generator Trellis 2. Trellis 2 has been the best open-source image [music] to 3D model out there, and it has been my go-to recommendation for months since [music] it was released basically. PixArt 3D is now built on top of Trellis 2 by Tencent, and it was just released under an [music] MIT license. So, fully

open-source, free to use, and we can even use it commercially. So, what makes PixArt 3D special? Well, the main difference to Trellis 2 is that PixArt 3D is actually able to read your input image way more accurate than Trellis [music] 2 is able to do that. And I'm going to show you this in a second, but

basically the front view of your model, now matches your input image almost perfectly, and this is of course great because we are always looking for the highest accuracy in our 3D models [music] that we're going to generate. But, here's the catch because the official PixArt 3D requirements say you need 24 [music] gigs of VRAM to run

this. And looking at the comments, a lot of you are running into exactly this problem. Some of you are even running out of system RAM on top of that. But, just like I showed you with Trellis 2, we are going to solve this with the GGUF quantization. [music] Massive shoutout here again to RR X who built the GGUF