30 blog entries for November 2023 -- as many posts as there are days in November!
This was a crazy month for AI as Sam Altman was fired from Open AI, and then rehired again. Plus there is a suggestion that Open AI has recently achieved AGI (and more scuttlebutt and doom and gloom).
Meanwhile I pushed my drawings further with AI -- through video, 3D and the improving 2D AI image generators.
END of Another ERA
We had the last drawing session at the NORTH FOURTH ART CENTER in Albuquerque -- forevah -- on November 16th, 2023, as the art center is closing the end of the year. The sessions started on December 2, 2021, and I first drew there on December 9, 2021. A big thank you to Athan (Tim) Psomas for monitoring the sessions for two years.
BUCKET EXHIBITIONS
something to show for it
At least we had something to show for this era. I had been bringing the Tortilla Press to the drawing sessions, so that during the model breaks we would make prints. This cascaded into the BUCKET EXHIBITIONS, and we showed many of those prints at the "Deck The Walls" exhibition at the North Fourth Art Center.
In fact, we showed 88 prints total -- from Albuquerque, Santa Fe, New Mexico, Texas/elsewhere and Mexico.
showing at the North Fourth Art Center
in Albuquerque
AI VIDEO
This seemed to be the month of AI Videos as I discovered new online programs and pushed the AI videos as far as I could:
- RunwayML Gen 2
- ModelScope Video2Video
- Stable Video Diffusion (on Replicate)
- Stable Video Diffusion (on Hugging Space)
- EzGif
- Final Frame
- Deforum -- we broke it so we could not use it this month
- Stable Video Diffusion (pinokio) -- we could not get it to work yet
I think the YouTube videos probably show off the video AI muscle better than any explanation:
I further altered the PlaiDay video output
with more AI using ModelScope Video2Video
They are now getting closer to making something like a movie with AI:
- YouTube (Matt Wolfe): Make A Viral MOVIE Trailer with AI
- YouTube (Theoretically Media): The Future of AI Video Has Arrived! (Stable Diffusion Video Tutorial/Walkthrough)
PIKA LABS upped their AI Video program and changed their name to just Pika on November 28, 2023. I signed up for the waiting list in order to try the new features.
3D AI
3D in AI is also making big leaps. I pushed these three AI 3D programs this month:
- CSM (image-to-3D on Discord)
- Genie by Luma Labs (text-to-3D on Discord)
- MasterpieceX (text-to-3D animation)
- Note: Latent Labs doesn't seem to work anymore on Discord
CSM
DOWNLOAD: "ropesculpture.GLB" (3.81 MB)
Tim and Adam at Quelab helped me 3D print the above 3D figure, as both a small one inch resin print, and (later) a large 7.5 inch PLA print:
GENIE
On November 1st, Luma Labs premiered -- Genie -- the text-to-3D generator on Discord; and I generated a "robot javelina" in Discord:
DOWNLOAD: Javelina_Robot.GLB (4.95 MB)
MASTERPIECE X
Later in the month I discovered Masterpiece X, which would not only generate a 3D file from text, but would animate it too:
DOWNLOAD: An_artist_with_a_bad_haircut_1.GLB (18.83 MB)
2D AI IMAGES
Leonardo.AI
I revisited Leonardo.AI, mostly because it won't censor my figure drawings (unlike Midjourney); it also lets me train models from my drawings; and it generates the best AI results in my figure drawing style. It's still free, though there is a paid plan (which would not accept my VISA gift cards when I last tried it a while back):
- I also created a new model in Leonardo.AI, from 40 of my best 2023 drawings so far -- 2023Drawings
Midjourney
Midjourney introduced "Style Tuning," so that one can steer the results more towards a particular style. But there is nowhere near the control you get when you train a model in Leonardo.AI:
AI EXHIBITION
I saw the most impressive physical AI Art show so far, by Jonathan Loth, at the North Fourth Art Center, down the hall from where we were exhibiting our 8x8 inch prints:
Jonathan Loth filled the walls of a whole room
LCM
AI in REAL TIME
In November a few programs surfaced that would create AI versions of one's drawing -- IN REAL TIME -- for every stylus mark one made on the computer screen -- Latent Consistency Models (LCMs).
I'm hoping that these programs will "correct" my drawings, as I draw the live model in real time. Finally I might be able to draw!!!
because I can't draw
The Collab LCM program worked, but their model "corrected" my drawings into anime:
I also discovered Drawfast to make random AI images:
I bought a Samsung Galaxy S9+ tablet so that I could draw the model at the figure drawing session, and have the Internet correct my drawing in real time:
- Krea.AI introduced a new REAL TIME AI drawing program, though we did not have access to it in November
RANDOM AI
It is just hard keeping up with all the new AI ART programs, and I found a few more in the last hours of November, which I listed in the Random AI blog post:
PRINTS
We showed 88 prints -- 8x8 inches -- at the North Fourth Art Center during their final Deck the Walls art show. Next year in 2024 the art center will be closing.
I also pitched having a linocut printmaking workshop in Santa Fe, in the classroom at the new Vladem Contemporary Art Museum:
QUELAB
I spend a lot of time at Quelab hackerspace in Albuquerque. I aspire to try to use all their tools to make something physical from my drawings. Adam CNCed one of my drawings into acrylic:
FIELD TRIP
SOFTWARE
- Midjourney "Style Tuning"
- pinokio
- ModelScope Video2Video
- Luma Labs Genie
- <model-viewer>
- EzGif
- PIKA LABS
- CSM
- Poly (for creating textures with AI)
- PlaiDay (allows face swapping in AI video)
- Moonvalley
- Melobytes (AI music generator)
- Petalica Paint
- Palette FM
- MasterpieceX
- Gaussian Splatting in Polycam
- Looking Gloves Video
- Krea.AI
- EverArt
- Leonardo.AI
- Hugging Face: stable-video-diffusion
- Deep Dream Generator
- Sketchbook app
- Replicate: lucataco / real-esrgan-video
- Civitai
- FinalFrame
- ClipDrop -- SDXL Turbo
- TLD Drawfast