I use Glaxnimate for SVG and Lottie animation: https://glaxnimate.mattbas.org

Tried it but it doesn't work on macOS arm64... :(

A Lottie animation generator that allows to create a Lottie animation from an image and text input.

That sounds really cool. How would you create such a thing?

The current workflow is create an image with stable diffusion, vectorize it, convert it to Lottie format, and than use a function that can create the Lottie animation keyframes.

How would you rig whatever is in the stable diffusion generated image automatically? You'd have to find a way to identify the objects in the image, segment them into their individual limbs/pieces and rig and animate them.

Maybe Meta's open source segmentation tool would be handy. Maybe you can rig the person/animal/object in the image using one of the object detection tools on HuggingFace.

This seems like a really interesting project. I'd love to check it out if you plan on open sourcing it!

Actually there are models for doing the layering.

1. https://github.com/SketchSeg/SketchSeg-Natural-Prior

2. https://github.com/IamCreateAI/LayerAnimate

3. https://github.com/hmrishavbandy/FlipSketch

Ich will hier weg

My theory on all the hype around agents is that AI companies want to establish themselves as the leading choice, so they can earn affiliate revenue for everything that gets sold over there agent.

I don't think there's profit in regular users in the forseeable future due to energy costs. The revenue will come from replacing you, not from helping you to buy a pair of sneakers.

Pretty amazing, two years ago it was basically impossible to recreate the same character in different poses.

My guess is that Google is trying to reduce AI spam and that your translation triggered something there.

for SVG and Lottie animation Glaxnimate: https://glaxnimate.mattbas.org

for simple pixel-based image animation/distortion Anime Effects: https://animeeffectsdevs.github.io

Yes, there is: https://alexk111.github.io/SVG-Morpheus/

Awesome, thank you!

It's definitely possible. The founder of Khan Academy demonstrated something like this with GPT-4o: https://www.youtube.com/watch?v=IvXZCocyU_M

Listen to someone who has worked in the industry: https://youtube.com/watch?v=lJGa3rtuoKY