Copied From https://tensor.art/models/835190737831415493
"
With just 4 steps and cfg=1, the original model's generation effect of 25+ steps can be achieved, and the generation time is only one-thirteenth of the original. It takes only 1 second to generate a 768x1248 image on a mid-range GPU like RTX 3080.
"
Description
v1
Details
Downloads
173
Platform
CivitAI
Platform Status
Deleted
Created
3/16/2025
Updated
10/16/2025
Deleted
10/14/2025
Files
tensorart_sd3.5m_4steps.safetensors
Mirrors
Huggingface (1 mirrors)
CivitAI (1 mirrors)
Other Platforms (TensorArt, SeaArt, etc.) (1 mirrors)