I've been able to get some okay results out of it -- touchier than the other two loras, it feels like, but there's probably ways to boost up the results with some preprocessing. Trying to find a good way to do a mannequin conversion as is.

File not Found

Hi. Could you please share your workflow?

Much improved results with 2.0. Vastly more likely to at least attempt to mimic the pose, and most of the attempts have good to great results. Nice.

wf?

any tips on stop it from transferring the outfit too?

Nice! I was having problems with some positions, but I found that describing the difficult positions make the efficiency of this lora goes up to 100%. Thanks a lot!

Не работает при сшивании

QWEN transfers poses well; if you add an image in openpose format and say, "Change the character's pose to this one," it transfers 1 to 1.

When I try transfering the pose, it matches the body size/proportions as well. Is there a way it can only transfer the pose? What can I do in cases like this? I followed the guide with the strenght/cfg etc.

Doesnt work. Just returns same input picture (stitched pose and character pictures) as output.

How is the split image source processed and then concatenated into three? Any special node process? Any required size / ratios? None of the examples have an embedded workflow, and there's no explanation here, or the linked webpage.

What's the process for the input dimensions? It would be foolish & risky to run a JAR file to do such a simple task. If you could be more open about how all this works / share specific details, that would be much more trustworthy. I want to try this, but I'll pass until the process is more clear.