KRЯRL DЯAWINGS

The challenge was RIGGING the 3D figures and animating them. MESH-to-MOTION is a free online program that does that nicely (HexaGEN claims to do this online as well, though I did not try it).

MIXAMO is good for animating 3D objects (import as OBJs), however I had problem with the output. They were not colored and textured, and they imported too high into the AR program, so that no one could see them.

I used BABYLON.js 3D Viewer to reposition the MIXAMO 3D models, and also to view 3D models online.

*****

Before animating, I converted a lot of 2D images into 3D using ComfyUI Hunyuan 2.1Image-to-3D Simplified. This program was free and works decently, though I think I get better results with a more recent version -- Hunyan3D 3.1 Pro -- in KREA.

The huge advantage is that I can use the simplified version of Hunyuan3D for free, on the powerful AI computer at Quelab. However that local computer does not have enough VRAM to generate colors and textures. I believe that we need 28 Gigs of VRAM to run such a 3D program locally.

I converted a lot of the AI ORIGAMI ANIMALS into 3D GLB files, as well as a lot of my figure drawings that had been enhanced with AI, on April 9, 2026. Again, there are 2D-to-3D programs that do a little better job.

Also note that MAKERLAB by Bambu Studio will generate a colored and textured 3D file from a 2D image.

Now perhaps I have to also generate my own motions with custom BVH files -- such as those that I can generated with (Hugging Face) MoMask. I blogged about this in January 2024 -- AI Confusion -- and ABQ Art Walk January.

(YouTube) Importing Mocap into BLENDER
(YouTube) Hunyuan 3D-2.5 - Create and Rig a 3D Character in 5 Steps

**********

I like breaking the 3D files into pieces, which can be done nicely with SegviGEN...but that program has been paused on Hugging Spac e.

However SegviGEN can be run locally -- but needs at least 24 gigs of memory:

docker run -it -p 7860:7860 --platform=linux/amd64 \
	registry.hf.space/fenghora-segvigen:latest python app.py

SegivGEN can be run locally

You need to switch on the new containerd engine in Docker for pulling and storing images. This will be the new Docker default in the future. Read more:

Extending Docker’s Integration with containerd

*****

Clone (git)

Running locally is potentially dangerous. Make sure to review this Space code before proceeding.

# Clone repository

git clone https://huggingface.co/spaces/fenghora/SegviGen
cd SegviGen

# Create and activate Python environment

python -m venv env
source env/bin/activate

# Install dependencies and run

pip install -r requirements.txt
python app.py

Also note that the segmented GLB files can be compressed still with GLB Compressor to very small files

Thursday, 9 April 2026

Snake Silk Screen with the Laser Cutter at QUELAB

Virgil Velasco sent me this image a while back, as I promised to laser cut it into silk screen. The image was too small, so I enlarged it with Deep Dream Generator. Then I laser cut it into silk screen at Quelab before the drawing session at Henry Morales' studio, on April 7, 2026:

OP ART SALAD

The snake drawing by Virgil Velasco

Image inside of 4x4 inches

On the laser cutter at Quelab

In the Lighburn Software

Speed 150

Power 30

The deeper settings that worked

I first laser cut into a larger metal screen. Though I did not print with this, it seemed to come out worse, as the laser discolored everything within the inscribed 6x6 inch square. I believe this was part of the settings in Lightburn (which I changed later):