Wednesday, 22 January 2025

BRAIN MAN

BRAINSTORMING with AI, but for 2025:

Driving into Albuquerque in the 60s
 from Santa Fe

January 21, 2025
Midjourney
Derived in Midjourney from this image
(--p m7269599726876491811)
(--profile z9nizie)

Midjourney
VIDU
EzGIF
BRAINSTORMING!


Midjourney
Viggle
My brain has a theory, 
and it's mine, and no one else's

DOWNLOAD:  Brainman.GLB  (1.61 MB)


DEEP SEEK
MULTIMODAL UNDERSTANDING

I asked the new Chinese AI open source program -- DeepSeek -- to describe my BRAIN MAN, by uploading it to Multimodal Understanding, listed on the DeepSeek Hugging Face page:

WHAT IS THIS?

Multimodal Understanding

"This image appears to be a surreal, colorful illustration of a humanoid figure with an unusual and fantastical appearance. The figure is standing in front of a house with a distinctive, whimsical design. The character has a large, elaborate headdress with what looks like a mix of organic and mechanical elements, including a large, round object that resembles a brain or a globe. The figure is wearing a patterned top, green pants, and has multiple limbs and appendages, giving it a surreal and otherworldly look. The overall style is reminiscent of surrealism and fantasy art."

Then I uploaded the above prompt to the same program -- Multimodal Understanding -- and asked it to generate images:

Images generated from the prompt
that was generated by Multimodal Understanding

Multimodal Understanding
One of the images generated by
the Multimodal Understanding prompt

I am a little confused.  I thought that the DeepSeek LLM was UNABLE to generate images -- so maybe the Hugging Space website is using an additional image generator of some sort to create images (but which one?).

I then uploaded the above image, which was generated in Multimodal Understanding, and reworked it a few times in Midjourney Moodboards:

Reworked in Midjourney Moodboards

I think that I can even install DeepSeek R1 on my laptop and run it locally.  While it will not generate images, it can generate code for programs that alter images -- I did that already with ANIMATION.

I am sure we could install DeepSeek on a computer at Quelab, and run it within the makerspace.  Ideally we would get a computer with a powerful Nvidia video card, and install AI ART programs from Pinokio to generate images -- like we did in August 2023 with the open source Automatic 1111-Stable Diffusion, and later generated Deforum AI videos.

IMAGE FX

UPDATE (January 26, 2025): I uploaded the same Multimodal Understanding prompt into ImageFX by Google, and it generated images similar to the Multi-Modal program:

Posted by at

No comments:

Post a Comment

Note: only a member of this blog may post a comment.

Subscribe to: Post Comments (Atom)