halutmatmul VS 3d-ken-burns

This master's thesis sort of does it, but it doesn't have any fine-tuning yet so it completely wrecks the accuracy: https://github.com/joennlae/halutmatmul.

If someone worked on contributing this to Composer [1] I'd be down to help out. I can't justify building it all on my own right now since we're 100% focused on training speedup, but I could definitely meet and talk through it, help code tricky parts, review PRs, etc.

[1] https://github.com/mosaicml/composer

I guess you could use a free video editor (Davinci resolve is free with less features than payed version). And then try to use this open source script: https://github.com/sniklaus/3d-ken-burns, but it would definitely be harder.

installing ROCm is bit of a pain (there is little packaging, so you have to rebuild it yourself)

Search who's running Stable Diffusion on Nvidia and who's running on AMD: if you are using AMD, you are kind of on your own.

Finally, you have model with custom CUDA code (e.g. https://github.com/sniklaus/3d-ken-burns )

There's a great Github with a script that I run on Google Colab that does a decent parallax effect in about 30 seconds through ML... but it just takes a while to get the instance spun up, and I've yet to figure out how to push out a 4k .mp4 from it. Surely someone has coding chops and can do this for parallax and monetize it like the Dall-E bot?

Making use of: https://github.com/sniklaus/3d-ken-burns