Happy to announce DreamFusion, our new method for Text-to-3D! dreamfusion3d.github.io We optimize a NeRF from scratch using a pretrained text-to-image diffusion model. No 3D data needed! Joint work w/ the incredible team of @BenMildenhall @ajayj_ @jon_barron #dreamfusion
@poolio @BenMildenhall @ajayj_ @jon_barron I'd be very interested to see what text was used to create these. I'm skeptical of much of what's touted as AI imaging.
@SlicinHammer @BenMildenhall @ajayj_ @jon_barron all results are in the gallery w/text: dreamfusion3d.github.io/gallery.html adding prefixes to the prompts like "a DSLR photo of" was the only tweaking we sometimes did
@poolio @BenMildenhall @ajayj_ @jon_barron All the descriptions are rather jargon heavy, but what I'm getting is the 'AI' is "pre-trained", which I take to mean has a selection of standard models to start with. I mean, if that's so, it's not really AI. That said, the ability for computers to do finish work is amazing...