Updated: Nov 30, 2025
base model
I just built a new workflow that uses only Z_Image, and the results are surprisingly good.
I used the same base as my SDXL/SD workflow
it generates much more variety in poses, compositions, and scenes using the same prompt.
The trick is simple:
- Two KSamplers, just like the usual setup
- But the first KSampler runs completely empty, low resolution, and only 2 steps
- This makes it super fast and gives Z_Image a rough base to reinterpret with much stronger variety
The second KSampler (Z_Image) then takes over and produces the final image with improved pose, framing, and overall creativity — all from the same prompt.
Last image is only Z_image for comparassion