Stable Diffusion vs. Houdini
The first experiments with the most promising Stable Diffusion implementation yet! The aim was to see how much consistancy one can generate with different promts based on the same highly detailed model.
Team
- Nicolas Lekai — CGI
Client
In-house
As the input for the SD model, I used the OpenGL depth map of Houdini’s viewport. - what we are seeing above is a ‘velocity field’. I was amazed by how much the AI engine was able to stick to the intricate details when generating it’s own ouput.
