ComfyUI-GLSL: run GLSL shaders as GPU-native image ops in ComfyUI, 4K in under 0.5s

itsAlright_its0kay · reddit · 2026-08-22

A developer open-sourced ComfyUI-GLSL, bringing GLSL shaders into ComfyUI workflows as a true GPU image-processing layer. The runtime uses Vulkan compute + GLSL → SPIR-V, processing a 4K image in under 0.5s on a 3090, with chainable nodes for advanced looks.

Key features:

GPU-native Vulkan compute execution, .glsl shader library + inline shader node
Automatic detection/adaptation of shader dialects (Vulkan GLSL, GLES/WebGL, GLSL Sandbox, Processing TextureShaders, GIPS)
Dynamic UI controls generated from shader uniforms, multipass GIPS filters, optional mask input
Pipeline/shader caching, discrete GPU preference, Comfy IMAGE tensors in/out

The interesting shift: instead of writing a Python node for each image operation, you can just write (or paste) a shader—shaders from other ecosystems often adapt to the Vulkan backend automatically. It ships with a sizable GIPS shader library (color, blur/sharpen, distortion, edges, effects). Still early; the author invites shader users to try and break it.

Repo: github.com/theRealAi/ComfyUI-GLSL

Original post →

ComfyUI-GLSL: run GLSL shaders as GPU-native image ops in ComfyUI, 4K in under 0.5s

More from coding & agent