Underview v1 - A Penis Model for Z-Image
This model is version 4 of attempting to train male anatomy with z-image turbo. I used the z-image de-turbo model provided by Ostris's Ai-Toolkit. After many attempts, I found that training at a high step count (this model is at 12k steps with 50 images at 512px with the de-turbo preset) and essentially over-baking allowed a low strength to be used so that quality of the image isn't badly degraded. This model is best used with a strength of 0.50-0.67, the quality should be similar to base (or better) and remains very flexible in outputs. It works best with a CFG more than 1 but less 3, I usually use 1.8 at 12-20 steps. You may need to do a low de-noise after to get some finer details. The model was trained with captions provided with the AutoDescribe-Images tool provided by hydropix on github. The trigger word is Underview, I specifically avoided training images of the top part of the glans as loras tend to get confused on which way is top or below, so keep vantage points at a lower angle and expect to get gens with the penis pointed up or to the side.
Underview v2
v2 is a modified version of v1 with an improved and much more consistent output. thanks to the amazing RealTimeLora v2 node shootthesound made!
After trial and error, the layers 24, 27, 28, and 29 could be turned off and eliminate most of the unwanted noise/artifacts that are present when the lora gets to full strength.
The lora also seems to understand higher angle views now, keeping it named "Underview" may be underselling it.
The RealTimeLora node pack also comes with a lora scheduling node that greatly improves the output.
The best schedule for the lora is (0:0, 0.2:0.65, 0.85:1, 1:0.65).
The v2 workflow is here: Underview v2 Workflow
The v1 version of the lora is still recommended to be used as a base pass (0.65 strength) before passing it to the v2 lora (with the above strength schedule).
v1 is better at following the prompted body types if you are wanting a less defined physique.
Description
v4 for me but v1 for release, focused on male anatomy from a view looking up
FAQ
Comments (11)
Those are some great quality outputs, would you mind sharing your workflow?
Very good quality! This is the best lora for this right now in the whole site.
Omg. THANK YOU.
This is exactly what I was waiting for. A lora capable of doing actual penises and not deformed meaty abominations. The best part is that it doesn't affect the original composition. No quality degradation or flux-like plastic skin.
Yes, I tried and failed for a while to get it to work at full strength, artifacts always started showing up before the model could learn regardless of training settings. I found that the higher the steps (and the more inflexible it got), the more knowledge the model kept at a lower strength, which ended up eliminating the artifacts that have been plaguing z-image lora training and prevented the base model from forgetting what everything other than a penis looks like!
@n8te Nice. Do you plan to do training with other angles or states (flaccid, erect pointing in other directions)?
@anthonyless Yes, I've got several datasets ready but I'll most likely wait for the full model release
I still cant find a LoRA for Zimage that can do flaccid penises...
@n8te base model has been released! waiting for an more amazing lora
Amazing! What model did AutoDescribe-Images get the descriptions from??
I used qwen3-vl:8b for the model and I also added specific directions to the config file. I am going to write up a guide with what worked for me that I'll share on here this weekend.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.