Mystery Lady - Exotic Dance Pose Helper
11th March 2023 Latest Update - Version 7 - 837 Poses
This Pose Helper is based from a large collection of many Exotic Dancers
This will be expanded in future.
other packs: PowerRangers, JoJo's Bizzarre, Gundam
Thanks for stopping by! (CLICK SHOW MORE)
More demo images and example will be added soon!
Use these in Control NET with Open Pose with any SD 1.5 model
video tutorial at the bottom
Optional: Includes each of the poses in JSON so you can alter them if you wish, once you have saved your edited Poses into PNG you can use those with Control NET.
Enjoy ! and post your images in the review ;)
futures versions will release with new Poses for this collection
(Poses containing high numbers of Persons are experimental)
release log:
Version 1 contains 24 Poses
Version 2 contains 94 Poses
Version 3 contains 202 Poses
Version 4 contains 394 Poses
Version 5 contains 602 Poses
Version 6 contains 774 Poses
Version 7 contains 837 Poses
Tutorial:
1. Have Automatic Web UI installed
2. Use the ControlNET extension
3. Download the ControlNET models as explained here to your WebUI
4. Choose a nice SD 1.5 model (suggestions at the bottom)
5. Restart the webUI and expand the Control NET Panel:
There are many ways you can use this tool, but for the OpenPose system, drag your pose in to the image space (1) then click enable (2) and finally select ControlNETPreTrained_openposeDifference (3)
From here you can use txt2img to prompt away like normal :)
Description
add 208 new poses - new total is 602
FAQ
Comments (13)
Is there a way to alternate through all of them iteratively with the same prompt? Like have a batch and it feeds a new pose every time?
i know people that have been working on an extension for this, once it is released i can add it to the description, until then we must do this manually :)
@muuuuuu What i do is add all the poses i want to a video using the poses as frame. after that load that video in the m2m with control net openpose selected. this will then make an image and then move on to the next pose for each frame of the video you made. it works great :) the extension i was waiting on never got released
That's a good and useful concept, but it would greatly help to see some renders made with these poses.
currently i only had the time to create the poses, my work is SFW, so while i can add some examples, your choice of model and prompt has a huge impact on what you will get. Some reviews were submitted showing this. If i get a chance i can add some skeletons overlaid in the description soon.
Thanks. To see the skeletons alternated with the corresponding renders would be of great help for all the users. Anyway, great job !
@ritcher1 in all cases these poses are based on copyright photos (for the best accuracy its all persons). I do not own the rights to those images for republishing, however i can get some public domain examples in V6.
V5 is releasing today :D I've been busy expanding the sets
Successfully applied the pose, but the face became weird. Is it possible to apply pose without changing face?
The pose should not affect the faces, as they come from your model first then the Lora / TI stack. I would look for face enhancing LORA/TI if you didn't have any or use a negative prompt embed.
Can still inpaint if the face is the only oddball. Upscale x4 and then mask the face with around 0.30 denoising 1024x1024, or 768x768 if vram is not sufficient.
Just generate 100 pictures. It is like gacha.
I think you can also use ControlNet from the image2image / inpainting tabs. idk how well it works but it might save the face better
@1_Boring_Cat_Rabbit I use "djx woman" (sd 1.5/2.1) to enhance faces (lora) but to be honest if you have a little denoise on hires (0.3 - 0.35) and use a good model it will enhance the faces for you in the hi res pass. even going from 512->768 is enough. The model will determine how good the faces will be tho. Lyriel is a good 1.5 model at the time of posting. I love inpainting faces but if i can avoid the step it saves a lot of time.