User asks about async VAE decoding to cut ComfyUI generation time

deepsky88 · reddit · 2026-08-23

A ComfyUI user notes that while Turbo LoRA and attention optimizations helped, VAE decoding still eats up about 30% of generation time. They are asking if there is a way to implement asynchronous logic within ComfyUI workflows to handle such operations more efficiently.

Original post →

More from Apps

Apps channel →