For controlnet openpose/depth.
Please read the version details for version-specific information.
Thigh focus: Better with depth
Looking/facing away: Better with openpose
Description
Edit: These require controlnet/openpose
6 different poses. 3 of these have the same base pose, just minor differences.
I attached a batch of images for each pose to avoid cherry-picking. What you see is what you get.
Please notice that changing key words in the prompt will change the angle significantly. For example, as the pictures show, pose 3 will produce a completely different angle with "from_above", compared to "from_below" in your prompt. So there is a lot of flexibility to go with these poses. (There is also "from_side" and "from_behind", though I use these less frequently.)
Poses:
1) This one works well with 512x512. They may be opening a box or something, at least that is what I was going for.
2) Works both with 512x640 and 640x512. Also produces some interesting results when you change the angle to "from_below" and use "crawling" instead of "kneeling" in your prompt, though the arms tend to be a little short in those outputs.
3) Works better with 512x640 or something similar when you're going for the "from_below" angle. If you go 512x512 or a wider ratio, the hip tends to push past the left arm. Changing the angle to "from_above" also produced some interesting results. This is the least consistent of the 6 poses in producing good results.
4) Pretty straight forward. More close-up than anticipated.
5) This one only works when you really emphasize in the prompt that you want a lying pose. Pretty consistent when you get there.
6) A selfie pose, fairly consistent. You may also go for "from_below" and "crawling" here to get something completely different where they are leaning on something with their arm.
FAQ
Comments (7)
Thanks for making and sharing these! It's so hard to get a good low angle or high angle shot. This will help a lot. ๐
I love these, especially the selfies. Although I've been able to create some of these poses--including the selfies--via prompts alone, it's only happened after wrestling it out of the models AND sheer luck, LOL. These strange-at-first poses just made it a thousand times easier, and (more) guaranteed. Like the others, thank you for the contribution to the Community.
BTW, may I know what model/s did you use for the suspiciously-Lara-looking character? (And it is pre-reboot to boot,wow!) I'm still really new to this Scene (just a couple of weeks old), so the answer might be really obvious to the jaded eyes of a Community veteran, but to me it's not , heh.
I'm so impressed there is no 'climbing up' (walking up) pose in the community yet. I tried to use 'climb', 'climbing up', or 'climbing a rope' tag in many checkpoints but the character doesn't understand.
I ran into the same problem. Even with creating some manual climbing poses and running them through openpose it took a lot of adjusting, photoshop and inpainting to get it done.
They can be a bit funky to get to work, but I added some climbing poses here: https://civitai.com/models/107427?modelVersionId=120142







