growth 2

 

Same 'growth' text prompting keyframe script as the previous post (fixed seeds).  Both of these are driving the generative synthesis U-Net latent from Studio Artist paint action sequence (PASeq) output.  So all of the movement in the animation is being driven by what is happening in the PASeq animation output.

growth

 

All use the same text prompting script.  First is interpolating the text latent vector (fixed seed), second and third are recursively feeding back affine transformed previous frame (seed different for each frame).

change 3

 

Two more generative image synthesis runs using the 'change' text prompting keyframes.  Above was generated using recursive feedback of an affine transformed previous output frame into the U-Net latent input (random seed).  Below is using interpolation of the text prompt latents with a fixed seed.

change 2

 

Using the 'change' text key-framing script from an earlier post with a Studio Artist flow transition context (derived from 50 news image frames with 10X time expansion) feeding the U-net latent input in the generative ai context.  Two different 'strength' settings for above and below (.3 and .2).

EBM Vector Multi-pen

 

Starting to dip my toes back into multi-pen paint presets.  These are vector EBM (energy based model) paint examples.

walkabout -test 4

 

This is using the static 'walkabout' text prompt (no key-framing) with the exact same Studio Artist gallery show MSG QuadMixer mutation output i used in the previous 'change' animation post feeding into the generative ai context.

Below is the Studio Artist gallery show session output i referred to above, that is being used to feed the generative ai context U-Net latent input.  The gallery show session was using the QuadMixer technique running a random walk through a 4 preset MSG parameter space mutation.

change

 

Gallery Show QuadMixer MSG mutation session driving the generative ai context with a key-framed story line.