Luma AI Text-to-3D Models Tested

preview_player
Показать описание
Artificial Intelligence (AI) Generated These 3D Models using Luma AI Imagine. This is text-to-3D Model

Today, I tried the AI 3D model generator from Luma AI. Let me show you two 3D models. With the text prompt: ‘a realistic Dutch bunny’ I got this: it’s great! The styling is good, the texture even has small little hairs. And most of all: it looks cute. Then I was inspired by other bobblehead examples from users, so I typed: “A bobblehead of Steve Jobs” and I was curious if the tool could get the glasses right.: But, as you can see: from afar it looks nice, but it has these hilarious eyes. Almost perfect!

Luma is working on the future of personal 3D model generation with their tool called Imagine. It’s currently in an Alpha stage, so don’t be fooled by the good examples like my bunny. It’s not finished yet. But even when the generated 3D models are not good, I am enjoying myself A LOT with the experimentation. And with the current advances in AI, I am sure it’s going to be improved a lot in the next few months. And I am very happy with this development, because these generated 3D models will look great inside your own AR tour, using Wintor’s no code AR tour editor of course!

The tools used in the video:
Рекомендации по теме
Комментарии
Автор

How long did you have to wait for your invite? I've been on the imagine waiting list for nearly a month now

ezearo
Автор

I'm curious what it could do with a (Image reference to 3D Model)

RpgBlasterRpg
Автор

How long did it take to generate the model? A few minutes?

NijelHunt
Автор

It's sort of surprising how far behind real 3d geometry generation is compared with raster generation from diffusion models. As this doesn't feel like a particularly hard problem compared with say MJ 5, I sort of suspect the lack of performance in 3d generation is because most AI researchers do not understand 3d geometry well enough to have smart ideas for an approach to ml model design, and I would guess specifically struggle for smart ideas for a loss function framework (probably the most critical), so they've probably only tried a million or so model ideas, rather than the billions of attempts at 2d that have led to modern diffusion techniques through much more trial/error.

googleyoutubechannel