MASTERPIECE X
prompt: "an artist with a bad haircut"
motion: walking
prompt: "an artist with a bad haircut"
motion: walking
The incredible thing with Masterpiece X is that it not only uses AI to generate a 3D file, but it also ADDS ANIMATION to the 3D file -- and the GLB format contains all that information: the 3D file, color and texture, and animation.
At first I was disappointed with the files that I generated, as they weren't of full body figures, like the ones in their library. Plus it took a lot longer to generate the second file (which is the first one I uploaded to this blog post) than to generate the second one.
However on second thought, the 3D files are actually pretty good. I can probably get full body 3D figures with a more specific prompt. The first few models are free, but then there is a fee to generate 3D models.
LUMA AI
So on November 2, 2023 I generated a "javelina robot" in 3D, using "Genie":
Download: "javelina robot.GLB" (4.95 MB)
ALIEN JAVELINAS
Then on November 15, 2023 I generated some "alien javelinas" with Luma AI Genie:
prompt: "alien javelina"
as seen in the Discord server
Luma AI generated 4 versions
of the "alien javelina"
I liked the 4th version the best:
DOWNLOAD: alien javelina__highrez__04.GLB (3.03 MB)
But the 3rd version might be easier to 3D print:
DOWNLOAD: alien javelina__highrez__03.GLB (3.10 MB)
CSM
CSM allows one to generate a 3D file by uploading an IMAGE to Discord. This has gotten a lot faster since I first tried it last August:
DOWNLOAD: "Notwhistlersmotherchair.GLB" (3.7 MB)
LATENT LABS
The LATENT LABS program, text-to-3D, does not seem to work anymore. However back in May I generated a javelina in OBJ format:
Download: "javelina.OBJ" (2.63 MB)
- Bambu is the newest and perhaps the best 3D printer
BRONZE CASTING
The next step is to 3D print one of the AI files, and cast in in metal.
We did this in 2012. I commissioned Alison Aragon to sculpture a portrait in Albuquerque. Then I took it to Tucson and had it scanned in 2.5 D by Align-Rite. Next we 3D printed it, and sand casted the 3D print at Xerocraft hackerspace:
I would like to do that again at Quelab hackerspace in Albuquerque, with one of the AI 3D generated images.
- We bought some resin for the 3D resin printer
- Rio Grande Jewelry Supply sells 5 pounds of casting sand for $50
AUGMENTED REALITY
We use the AI 3D files in AR (Augmented Reality). The GLB files would probably have to be decimated first, to reduce the file byte size, so that they would load onto a smart phone quickly.
We can use WebAR, so that the viewer does not have to download an app (but the code has to be housed on a server somewhere). One can upload the GLB file into <model-viewer> to generate the code and test the WebAR, as we did for the 3D AR tapir below:
FURTHEMORE one can even take a snapshot of the half-real/half-AR scene, and even film it. We took a snapshot after uploading the tapir GLB file into <model-viewer>, and then animated it with AI using PIKA LABS:
In Short: We can use AR to MAKE VIDEOS WITH AR using AI:
- By uploading a still snapshot to PIKA LABS
- Or by uploading a video to ModelScope
3D GAUSSIAN SPLATTING
Also recently GAUSSIAN SPLATTING has surfaced, to make 3D scenes -- which we should be able to populate with 3D objects.
POLYCAM allows one to make 3D scenes with a series of photographs in 360°, or with 360° video. I tried it and the software worked -- but my scene FAILED EPICALLY:
I think I failed because I uploaded images of a turntable figure with no background. I suspect that the software needs some PARALLAX in the background to work to create a 3D scene...like with previous photogrammetry software, such as 123D Catch.
Still I should be able to make a good 3D Gaussian Splatter scene if I can upload images with parallax. But it's harder to actually use that 3D scene:
With an Add On we might be able to make animation in Blender, so the 3D Gaussian Splatter could be useful for making a movie (Unreal Engine is probably better for this).
HOLOGRAMS
We can even convert the videos made with 3D AR into holograms for the Looking Glass Portrait display -- using LOOKING GLOVES VIDEO.
While the example below did not use 3D AR, it did use AI to generate a video, which I then converted into a "quilt" to be seen as a hologram in the Looking Glass Portrait display:
DOWNLOAD: "PlayDay_AI_Quilt.JPG" (4.47 MB)
No comments:
Post a Comment
Note: only a member of this blog may post a comment.