GenAI News q4.23: More Insane Generative AI earthquakes

genAI Art Factory: Robots painting pictures

I’m just going to knock these out without much editing. These are the must-see tech developments of genAI content generation for Q4 2023. Are you tracking with us? Because according to our very aggressive / optimistic AI progress forecast 2022-2030 which we authored eons ago in Nov 2022, we’re already at about 2029 on the timeline…:


Pika Labs: Text-to-Video creation engine

Pika calls this “Idea to Reality” rendering, and that’s not far from the truth.

Luma: Text-to-3D model generation

The basic idea: type in anything (just like generators). It automagically creates a 3D mesh model, fully textured. A-fucking-mazing.

Try it yourself on Discord:

Important post-script: you can just as easily take these output meshes, and send them to a full-color 3d-printing service, to get your creations shipped to you in physical reality. #mindbender

LCM-Painter: Real Time sketch >

Here’s how it works: You put in a basic text prompt (key direction). You then live-draw / paint on the right. In real-time, a series of ever-morphing AI-Art renders is generated on the left. As you paint, the images morph and shift to match your color/shape direction. It feels… insane.

realtime genAI

link (free online demo):

Animate Anyone: Transforming Still Images into TikTok dances

This demo purports to do a number of things, all at once (three magic tricks in one!):

  1. Take any still image (either a drawing, a photograph, or an AI-generated still image).
  2. Feed into a video — of anyone or thing — moving
  3. It will translate the video into a dataset of skeletal movements — including fingers!!
  4. (MOST IMPRESSIVELY) — It will generate a full animation of the character you supplied in step 1, moving flawlessly mimicking the video you supplied in step 2. Including, purportedly, flowing fabric, reflections and shadows.

I’ll let the demo speak for itself:

Hmmmmmm. I have to say, this most impressive example — “Animate Anyone” — is highly suspect; no code has yet to be released, just that incredible, unbelievable (in the literal sense) demo video. Nonetheless, this GenAI feat undeniably WILL come to pass, and Adobe and many others are pressing hard to make it a reality.

The authors of Animate Anyone haven’t posted any code yet, which makes their “accomplishment” suspect. There are, however, many other examples of full codebases that you can try out for yourself. For example:

Concerned about Creative Jobs? You’re Not Wrong

Can you see what’s happening? Media Product (imagery, 3d models, animations) that used to take highly talented teams months (and millions) to create, now can be created by anyone with a smartphone and an imagination. This could very well mark the beginning of the end of the story of “commercial artists being able to make a living doing what they love.”

Thankfully, the Hollywood Writers Guild has done their best to get far out in front of this.

Read about how the Hollywood Writers Guild agreement with the major studios might just be the precursor / prototype for global UBI (Universal Basic Income) for the 20-80% of the workforce expected to lose their jobs to machines in the next decade.

DALL-E3: TEXT rendering for your genAI.Art

DALL-E3 has pretty much nailed accurate rendering of text in its images, whether it be thought bubbles, captions, labels, or t-shirt slogans. For illustration purposes, this is a game changer. Example:

DALL-E3 creates a cartoon with text on the fly

(nobody ever said that AI respected copyrightholy god!)

so this is all good entertainment.

I wonder how much it actually costs?

I am not alone:

P.S. : How much Energy does it take to generate an AI-Art Image?

(Spoiler Alert): about as much as it takes to charge a modern smartphone (for instance, iPhone 14, with a 4,000 mAh battery). The miracle here is not how much power it takes to generate an AI image, its how infinitesimally small the power drain is to fully charge a smartphone. The answer is approximately 0.004 kWh (kilowatt hours), which is far less than a penny ($0.01, or 1¢) on your power bill. amazing.

Of course, this doesn’t help slow the ever expanding juggernaut of global AI energy consumption…

article in MIT Technology Review:


prompt: make me an image: 100 humanoid robots in a factory, each with an easel in front of them, painting pictures. conveyor belts between the columns of robots ferry the finished paintings to mail trucks. put a bright LED sign hanging from the ceiling, with glowing red letters that read: “11.8 trillion served…”. at the bottom of the sign, in smaller letters: “…and counting”

engine: DALL-E3, OpenAI
— image editing sw used to manually crop to square, and add atmospheric fade to white.