Workflow to use SDXL/SD with Z_Image

This improves the variety you can get from the same prompt when using Z_Image.

I’m using an AIO model for Z_Image (download link included in the workflow).

I also included the nodes for the default model, CLIP, and VAE in case you want to switch back.

With RTX 4070 12gb vram inference is pretty fast with both models

This works for NSFW content as well. Z_Image still deforms genital areas a bit, but once NSFW LoRAs or fine-tuned checkpoints are released, it should work perfectly.