This LoRA was trained using both the Ref2VA and FL2VA models, then merged into a single release designed to help get that dick into the right hole.
It is intended primarily for close-up shots where penetration needs to remain clearly visible. It was trained to work whether or not a penis is already present in the reference image—but it still needs an adult bottom and a visible hole to target.
How to Use
Only one reference image and a prompt are required. Check the showcase examples for prompting guidance.
For my prompts, I generally use Qwen 3.8 27B Uncensored through an Ollama node in ComfyUI to expand the basic video idea. The resulting prompt usually benefits from some manual pruning: describe less of anything the model overemphasizes and add detail where it needs more direction.
It is as much art as science—but this LoRA knows how to find its hole.
Recommended Settings
No trigger word required—just a hole.
No Turbo generally produces better results.
Not sure how, but 5090 was able to use the 66GB bf16 file and generate in 4 minutes - amazing quality
Use a close, readable angle when penetration visibility matters.
Avoid excessive prompt details that compete with the main action.
If your reference image detail is lost, lower the strength of this LoRA.
I love the new latent upscaling workflows, but they did not work well with this LoRA. Some turbos wrecked em (pun intended). Splitting the sigmas sometimes caused the model to invent a new hole in the scrotum and penetrate that instead.
Optional Penis LoRAs
This version should generate a partner even when your reference image or video does not already contain one. For additional control, you can pair it with one of my penis LoRAs:
See the showcase for prompt examples. Most showcase MP4 files also contain embedded workflows that you can save and open directly in ComfyUI.
Please share your results using Add Post below!
Remember: 2,000 Buzz is worth about $2, of which creators receive roughly $1. These models require a great deal of work, testing, and electricity. Your support keeps me motivated to continue this second full-time job of mine.
Description
v1 merge of different checkpoints
Comments (4)
After the update, do we no longer need to make another purchase?
Another purchase wouldn't be required (the "during this time before the v1 release" note I believe just refers to the discount, as in those who jump in now get it at a lower rate), I usually pick up coachbates' early release models, and I don't recall ever seeing him charge again for the final release.
Since Civitai's purchase model means you're purchasing the "version" as far as I know, I believe he just renames the preview to v1 or whatnot so that people who picked up the preview retain access to v1.
A completely new release would (like a v2, or a new variant like the XXL release on his arousal LoRA), but I really don't mind paying another 50 cents for a new version/variant.
(The crypto deposit thing is more the annoying part, but the way I see it, the small bit of extra time it takes to go through that is a rounding error compared to the time and effort that coachbates puts into his models :D)
@Jumbo0290 Correct. No one purchasing a preview release will have to buy the v1 release - I just rename the version and you can download/generate with the new version. I've only had to split one up because the concepts were bleeding together, and then I reduced the price to half of that. Several people turned off by 1k buzz until you explain that is just $1 US
@coachbate Appreciate the confirmation! I feared a few minutes later after replying that my phrasing might've sounded far closer on the absolute side versus the "based on my observation" that I was going for - didn't want to speak on your behalf!
As always, I appreciate and love your work - we need more folks like you!