santa hat
deerdeer nosedeer glow
Sign In

Tutorial: Make your own ControlNet Poses FREE! (without a Gpu!) [REPOST]

Tutorial: Make your own ControlNet Poses FREE! (without a Gpu!) [REPOST]

Tutorial: Make your own Poses FREE! (without a Gpu!)

DISCLAIMER: This was written last year. You CAN use this tutorial to help yourself learn OPEN POSE via A1111 regardless. Some of the gradio spaces on Huggingface MAY not work even after i've duplicated them.


Warning: This tutorial is simple and written on the fly, there is no PDF version and the download is just the zip file full of the same images you see in this model card. Of the exception there ARE poses added to the zip file as a gift for reading this. (it wouldn't let me add more than one zip file sorry!)

This is an absolutely FREE and EASY simple way to fast make your own poses if you've unable to use controlnet pose maker tool in A1111 itself. This uses HUGGINGFACE spaces, which is 1001% FREE if you're using the spaces that are linked in this tutorial. You do not even need to LOG in at all.

Your first step is to go to: https://huggingface.co/

Once you're there you're going to have a CHOICE to login, or if you don't have an account it isn't required but it is helpful to sign up. This allows you to bookmark or duplicate spaces at will.

Again it's entirely up to you, this is what it looks like when you're NOT logged in, from here on in you'll see "DARK MODE" as that's our standard choice for websites.

Your search bar is at the VERY TOP, as you notice i'm logged in, so it's in dark mode. search for "POSES" - you'll have a few options show up.

You're looking for the two options of "JONIGATA" - One of them WILL say when you get it to it not to use it, but this is weird - in our opinion, because the older repository/space is a little better at recognizing line art/manga for poses (it's hit or miss but it works!).

If easier the link to this space is here: https://huggingface.co/spaces/jonigata/PoseMaker - Duplicated here: https://huggingface.co/spaces/EarthnDusk/PoseMaker other option https://huggingface.co/spaces/EarthnDusk/PoseMaker2

When you go to POSE MAKER (not posemaker 2) - it looks very similar, you'll notice this when you see pose maker 2 in a bit. You have options for width/height but we're focusing on the "ESTIMATION" section on both versions. Clearly you CAN edit the skeleton on both versions, as the command keys by save show in this screenshot.

In this example we spent time back in Second Life filtering through animated poses. We graciously resized each photo ahead of time to 512x768 - We didn't remember until it was too late that Control net has a maximum input size on A1111. It's best your images are smaller so you can use BIRME - birme.net/ for better resizing needs fast. You'll drop your resized or non resized image in the "ESTIMATION" box on the left hand side.

I'll show what your button options are and how to use them:

Once you've put your picture in, it'll tell you what the size of the picture is, and if there's an indvidual OR more than one in the picture. Once it loads this information before you do any estimation buttons, make sure you've pressed APPLY SIZE. This is important because you may not have the room/screenspace to resize it to the box it gives you.

"REPLACE" is where it replaces the default skeleton with the one that it's currently estimated from the picture. Provided you've already APPLIED the size at present to the canvas, you can replace the skeleton and either prepare to edit it or other actions.

The keyboard actions above are the ones that you'll be given to be able to edit before you save - don't forget you can just use your mouse to edit and create a more stylistic pose in case it missed a few details. Remember: This isn't perfect, you may have to stretch certain limbs to suit your needs.

Once you're entirely happy with your pose, hit save - this will save an image to your downloads folder, and you can then collect these and release your own pack OR HOARD THEM!

BUT WAIT THERE'S MORE! - We'll go through Pose Maker 2 - just to give you a 2nd go at this information!

Posemaker2 is a little more accurate for realistic (or video game) poses, though it can mess up sitting and/or half body poses. SO both pose spaces are REALLY good for different things. Most of the key commands are the same, and the estimation and edit process is the same.

IF easier than searching the link to this space is here: https://huggingface.co/spaces/jonigata/PoseMaker2

In case there were MORE commands (and I think there are) these are for posemaker 2, and allow you to edit your skeleton after you finish your estimations.

So again, this is an estimation but this time it's in Posemaker2 - you can see that the layout in HF gradio spaces is quite similar to the one you'd see in A1111. This may not be as simple for SOME, but if you're limited by space in your notebook this is fairly fast and quick and easy.

For example these are images used and the images output from the poses:

Both of these were based on an avatar we'd pre made over the years in Second Life. Second life is not related to Stable Diffusion other than we use it still for our needs and to make poses and make pics to make into Lycoris/Lora!

Here are example poses that are included in the pose pack that's included in this tutorial for fun!

So now you can see how to create your own poses, I'll likely cover some more gradio spaces on hugging face for civit users in future!

And yes, in the downloads are the POSES and the screenshots from this tutorial, but NOT the images used to MAKE the poses. You CAN re feed it AI created pictures to make poses btw!

28

Comments