Veo 3 now supports start frames. now you can create Veo 3 videos using any image as the starting frame! try it now in Krea Video.

New blog on my thoughts on research. I will have separate posts on Krea 1 model development and how I see generative model scene evolving. re-n-y.github.io/devlog/ramblin…

looking at the replies to this and thinking about how artists tend to feel strongly about ai because it implies a seismic self-narrative / identity level disruption to sense of self … this will happen to other fields too (see: people feeling grief and conflicted over AI winning

one of my proudest achievements is that i got a green card as a venezuelan under the trump administration

Mari When implemented in a reasonable amount of detail, things never work quite like you abstractly imagined they would. For this reason, prototyping helps you find the territory where you can have the best ideas. If you become invested in a bunch of supposedly-specific ideas before

Happy to have released my first model. A lot of work went into making Krea 1 open source.

Really nice write-up about post-training diffusion models!

WAN 2.2 image generation + a basic 4K GAN upscale is absolutely stunning try the combo KREA AI. prompts in alt

Speed up dit pretraining by changing your timestep sampling over time - start biased toward high noise for quick early convergence - gradually shift to uniform to ensure the model performs well across all steps In the first animation, you can see the train timesteps shift over