🛠️ A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)

Summary
A 2-person lab trained a text-to-image model using an efficient pixel-space diffusion architecture called Pyramid-JiT for roughly $20K. They are releasing the weights under an Apache 2.0 license as a stepping stone toward cheaper generative video models.

Why it’s interesting
It demonstrates an efficient pixel-space diffusion model trained for a relatively low budget of $20K with open weights released to the community.

Target user
Animators and creators interested in generative AI tools

Source metrics: Points 9 · Comments 0
HN discussion · Project
Source: #HackerNews / Show HN