Stable diffusion dreams of steam punk neural networks

A stable diffusion dream. The prompt was "ultrarealistic steam punk neural network machine in the shape of a brain, placed on a pedestal, covered with neurons made of gears. dramatic lighting. #unrealengine"
the new and improved v2 version of this video is now here: • Stable diffusion dream...
generated with this hacky script: gist.github.com/karpathy/0010...
The script slowly meanders through noise space to explore the space of possible generations for the fixed prompt.
Stable diffusion: stability.ai/blog/stable-diff...

[@michaelfischer841]

thanks for doing a literal stylized neural net cgi animation . brilliant !

[@meyklee]

Amazing video about this cool cyberpunk!

[@carykh]

so cool - imagine if someone really built a brain that was computationally the same as a real human one, but out of gears!

[@SahinKupusoglu]

Hypnotic... Had it ended with a fall, it would be just like one of my dreams!

[@2ndfloorsongs]

The diffusion in my dreams is never that stable. Also my dreams always seem to end with a hungry cat standing on my face. At least with cats, I'm no longer afraid of strange noises in the dark because they always have a ready explanation... This probably makes me susceptible to being eaten by monsters under my bed.

[@beaumontluke]

This just works so well

[@jonathansum9084]

Interesting..😆 I hope I can reproduce it for my own prompt.

[@2ndfloorsongs]

Can stable diffusion work with music sampling? Cohesive soundtracks with these videos would/might be cool.

[@evanacharya7981]

How are you traversing the latent space? Prompt variants alone have not provided such smooth interpolation for me.

[@DrKnowitallKnows]

Awesome! Is this something built into Unreal now?? Or is it a Dall-E 2 type engine with continuously varying inputs?

[@hayden-donnelly]

I think the unreal engine tag was just part of the prompt.

[@stuartmarsh5574]

very cool.

[@2ndfloorsongs]

The soundtrack from blueberry spaghetti would go really well with this one, or would that be too derivative?

[@HarblesTheSkeptial]

Would it work with Music? Or sounds in general? Train on the complete Spotify library plus the Library of Congress audio collection.

[@2ndfloorsongs]

My thoughts exactly, I hadn't read your comment before I wrote mine. Maybe it could be adapted to make VR motion that is less nausea inducing.

[@maxwellclarke1862]

Tip: when uploading this kind of thing to UA-cam, let UA-cam do the compression / video processing for you. Just make a really big, lossless video file and you'll get the best viewing results.

[@AndrejKarpathy]

i learn, thank you!

[@danl.4743]

@@AndrejKarpathy Tip: when writing I learn, use capital I. Or simply let UA-cam/Google autocorrect change the i to an uppercase I.

[@InaayasPlayland]

Dan L 🙈 you are crazy 🤪

[@Tubernameu123]

"pay me back" by using the knowledge to build something great" Some of us are not and may never be capable of building great things. We can argue or debate, "why" or the faults and failures. Or we can just accept it. Sometimes financial participation allows the lesser person to have a greater impact.

[@deeplearningpartnership]

Cool.

[@JulienReszka]

Cool

[@channelname501]

"Dreaming" is the perfect word to describe how I have reacted on the ChatGPT output. It was absolutely fake facts but with a great confidence

[@priyavratmisra]

1:05 I see C-3PO

[@reinerwilhelms-tricarico344]

What is it smokin?

[@udunus3874]

🤓it is more like clockpunk than steampunk🤓

[@PeterBaumgart1a]

Nice. But 2x too long.

[@danl.4743]

Yes. I agree.

[@TaliwhakerRotmg]

@@danl.4743 Why are you being so negative get outta here