How about some syntax like lh_ventral_down_thumb_down, rh_dorsal_up_pointing linked to clear non-conflicting representations?

ran this by chatGPT- it says sign language would be a good place to start checking for descriptions of hands-

It's an interesting idea to have a universal syntax to describe hand positions, especially for use in animation and CGI. This would enable a common language to be used across different software programs and platforms, making it easier to share and collaborate on projects.

The syntax you have proposed seems to be a combination of abbreviations for the left and right hands, the position of the hand (ventral, dorsal), the orientation of the hand (up, down), and the position of the fingers (thumb down, pointing). This type of syntax could potentially be expanded to include other hand positions and movements, such as gripping, grasping, and waving.

However, it's important to keep in mind that different cultures and regions may have different conventions for describing hand positions and movements, so any universal syntax would need to be carefully designed and tested to ensure its usability and effectiveness across different contexts. It may also be helpful to consult with experts in the fields of sign language and gesture studies to ensure that the syntax accurately reflects the nuances of hand movements and positions.

if this can be used to "repair" hands via inpainting, then yes, please do continue :)

如果能用来修复手就好了

full body下用这个lora大概率手是畸形的。所以还是不能用来修复或者绘制好看的手

只能用于这种展开的手势么，其他的比如半展，握住等可以么？

It seems, no matter what I do, all the hands that look promising during generation end up looking very skeletal - bony and pale. Has anyone else encountered this behaviour?

really nice hands, truly, but....we need them on a body somehow....maybe we could use this for editing the hands with controlnet?

Just like other LORA and EMBEDDING products that are said can be able to repair hands, a small amount of use can have a small impact on the hand, and the hand has a chance to be corrected.

When I adjusted the weight to 0.5, my images quality began to appear blurry and poor.

It's best not to have too high expectations for hand fixing. Perfect hands basically still depend on the prompt settings and your luck. Whether the hands are perfect or not basically has nothing to do with the weight you adjust.

Hi thanks for uploading the LoRA, could you please share more on the parameters you used to train and dataset? I'd like replicate your results and having trouble so far. Thanks!

this lora affects the whole image, removes the background. Why?

thx hand model!!!