This allows you to add some D to your text to image prompts, or your image editing prompts. I think the quality is pretty good. I created a prompt enhancer for my showcase images. Please share your results.
Created with my latest 200 image dataset and a few higher resolution images since this model's native resolution is 2048x2048 using AI Toolkit.
This is the first penis lora for this model that I know of, and looks like it is the first lora of any kind shared here on civitai for this model.
I found these to work better than the default settings in the initial workflow released that uses the dev (turbo/distilled) model are 10 steps (much less than recommended 28, which gives monster skin):
euler_a / sgm_uniform
euler_a_cfg_pp / bong_tangent
Workflows can be copied from the showcase image metadata
Description
30k steps
Comments (4)
Congratulations to the first ever HiDream O1 LoRA uploaded to CivitAI.
i'm finding the image editing with it quite good, and more consistent than text to image so far
how do you find it compares to Klein?
jury is still out - looking for loras that make it listen to my prompts for man on man action and frustrated left and right, which is why I started making my own