A demonstration of the new Stable Diffusion (sd-server) integration running within a headless Vulkan renderer. This pipeline uses img2img on original video source frames to generate AI textures, processes them through custom shaders, and writes the final output directly to disk. Optional CUDA acceleration is supported. The goal isn't realism; it is built strictly to generate highly randomized, abstract textures that work perfectly for glitch art and custom shader development. Shaders are applied after the texture generation for a complete automation of the process.
Note: The encoding process was speed up for this video because you wouldn't want to watch a terminal scroll. The process actually took a while on my older system with an RTX 2070.
Note: The encoding process was speed up for this video because you wouldn't want to watch a terminal scroll. The process actually took a while on my older system with an RTX 2070.
- Категория
- AI Картинки
Выполните вход или зарегистрируйтесь для отправки комментария.
Комментариев нет.
