media art et al ✨ half desi half modge podge
building /tmrrw

Here with popcorn to see what comes next 🍿

Update: I have a working GUI, we are at 2%, and have our first output

It’s accidentally using CPU so I think it’s going way slower than it needs to but w/e & I’ll try properly tomorrow

I have high ambitions for specificity but even if this doesn’t work out my understanding is expanding in a way that feels exciting!

✧the dev for onchain music
✧✧✧✧✧✧✧

Holding me accountable with prose & I respect that 

I’m interested in the digital debris of process so this model may yet come of use but next steps:

1) do model training without adding complexity to footage, more aligned to the actual footage I’ll give it, try more epochs

2) abandon pix2pix and explore controlnet combo for better(?) segmentation 

3) launch site that will be home for this eventually 

4) img2img translate screen recordings/recordings of self in process with the abandoned models

So you know, small and manageable, easy next steps

(For context color works but shape completely fails to translate)

these early sets are 
where we see if this is a 
flash of an experiment or
the seeds of a new practice
or maybe a mix of both