I made a sprite animator using Google's Gemini 2.0 flash experimental. Here's how it works: 1. Gemini 2.0 exp generates the images 2. They are uploaded each in an S3 bucket/folder 3. Then you can use WebGPU /w canvas or just img Let's watch some examples and generate some:

This is actually one out of many iterations, in the other ones you could crop/decide and add the frame, auto-generated dungeons and so on. e.g. I could also imagine another vision model segmenting the characters out of each frame or sprite-sheet

gg @GoogleDeepMind

Can Machine Learning beat the market? Check out this post on my free Substack where I share code and commentary for an XGBoost model and a Random Forest model that both deliver powerful performances.

If you have any fun features in mind, drop them below, the more crazy and "unachievable" the better

so clean

thanks, the goblin dancing and the flower ones are crazy

I wrote a script to do the same thing . UI is much better , Really cool

amazing; there's much to do to it, caching CDN etc and the UI can be even better too but very easy to develop I bet it can have great results with realistic images also, mini-veo 2.0

so cool holy shit

early access?

whoa this is so cool

thanks a lot, can early-access you too, I think the sprite/8bit stuff it's fun but imagine: 1. being able to control the sprites e.g. assign keys to each one 2. realistic images (easy) 3. many more video is just images after all and with the edit feature you could do a lot