The features continue to roll out for Luma AI.
Today Luma is unveiling a brand new product vertical with their research preview, Genie. Genie is able to take a simple text prompt, just like how you would with Midjourney, and from that can generate an interactive 3D model.
We've seen generative radiance fields begun to be released throughout the second half of this year, with various offerings including Dream Gaussians, Magic3D, and Luma's original method, Imagine3D.
From my understanding this is not a sequel to Imagine3D. Genie is a a research preview, that is not production ready, but shows implications about where text to full 3D is.
I've been fortunate to have been testing Genie for the last week or so and some of the outputs that result are just plain shocking. It currently is creating models within 10 seconds. And that's not just one output. It's four, fully interactive models that can be refined. Furthermore, users can play around with the material texture map and export to .glb.
It's clear that a decent members of the Luma team has been working on generative applications, in addition to Luma's main offering. Though Genie's main focus for the time being is for research and prototyping, the results are fast and surprisingly high quality. Now, where have I heard that before?
Luma has clearly also heard the positive feedback from their Magic Reveal feature and are stepping it up a notch, with a brand new visualization.
Similar to the rest of Luma's offering, it is 100% free and down the road is possible to have the resulting models be open sourced.
Given the amount of free tools that Luma offers and their associated backend costs, one can only imagine that Luma is beginning to look at a raising a Series B and given the reach of their product offerings, I imagine that it might be comprable to RunwayML's recent raise.
For long time, people have been speculating when Midjourney will bridge the gap into generative 3D; it starting to seem possible given this pace, that Midjourney will not be the first to ship.
