Holly Grimm started off the year giving an online live stream presentation of her AI art for Currents New Media, using her figure drawings for a database:
Holly Grimm transferred the Tuesday Night Santa Fe drawing group to Zoom, thereby saving it during Covid. She is also at the forefront of AI art. I have been following her example, trying to apply the AI tools to my figure drawings.
- Later on November 4th, Holly Grimm showed one of her creations -- GeoMorph #3, an AI-Generated Landform -- in NYC at the "Dreamverse Terminal 5" event/exhibit.
DEEP DREAM GENERATOR
A) Upper Left, the original scanned drawing
B) Upper Right, that drawing altered in AI by Petalica Paint
C) Lower Left, that drawing altered in AI by Deep Dream Generator
D) Lower Right, the two AI results blended using Layers in Photoshop Elements 2021
- Myoats
- Bomomo
- Weavesilk
- Mr Doob/Harmony
- Fluid Painter
- Verve Painter
- Dreamlines
- Christmas Experiments
- Programs listed from 2020
- May 19, 2018
- Computer Assisted Drawings (June 30, 2019)
- AI Art Review (December 29, 2019)
- First I seeded the VQGAN+CLIP program with the same image, but used various text prompts.
- Then I played "telephone" and seeded the VQGAN+CLIP program with the previous image result:
- I destroyed a figure drawing (after noodling it to death at a coffee shop)
- I uploaded it to Deep Dream Generator, and applied their in house "Deep Dream" style to it (top image).
- Then I uploaded the result to a VQGAN+CLIP program, and used the text "Deep Dream" prompt to alter the original Deep Dream image (bottom image):
Blog Entries:
- AI Art -- VQGAN+CLIP
- More fun with VQGAN+CLIP -- Sculptural
- August 24, 2021 -- VQGAN+CLIP -- Abraham AI
- September 7, 2021 -- Spatial
VQGAN Programs:
- CLIP & gradient ascent for text-to-image (Deep Daze?).ipynb
- Generate images from text phrases with VQGAN and CLIP (z+quantize method with augmentations)
- VQGAN+CLIP (z+quantize method with augmentations)
- VQGAN+CLIP An PUBLIC COPY
- VQGAN+CLIP (with pooling)👍
- This is not DALL-E
However no matter how well the AI process worked, it always lost all but a trace of my input drawing. In other words, the results were not really my art. In fact, there is a danger that everyone's VQGAN+CLIP art looks the same.
StyleGan2 Pretrained Model
- Plus my drawing model is mentioned on GitHub by filipecalegario/awesome-generated-deep-art
ARTBREEDER
I used ArtBreeder to make:- landscapes: A and B
- faces
- faces from my portrait drawing with a VQGAN+CLIP assist
Later I pushed ArtBreeder a little further:
DEEP FACE DRAWING
Monster Mash
MONSTER MASH is a simple online 3D animation software. Google has an entry about Monster Mash on it's AI blog.
Text-to-Image
I experimented with Hynogram and other text-to-image programs that blossomed at the end of 2021:
- Online: Text to Image generator
- improving of Aleph2Image(delta): CLIP+DALL-E decoder.ipynb
- Generate images from text phrases with VQGAN and CLIP (z+quantize method with augmentations)
- Generative Engine by Runway ML
- Kapwing text-to-video
- A Hugging Face VQGAN+CLIP
No comments:
Post a Comment
Note: only a member of this blog may post a comment.