My Apologies! Rosebud AI was free and education focused when I linked to it, I had no idea how much things had changed.
Newest Update: H42 Version, Using a resource I found on Civit AI to use with Openpose and IP adapter to make better character Sprite Sheets. If you want to tip for the H42 version, consider splitting the tip in half and giving a portion to https://civarchive.com/models/56307?modelVersionId=63973 as their resource came in clutch for testing this and saved me a lot of time lol. Received a slight update after rearranging background removers.
Rosebud AI Version. I use Rosebud AI to make games with my son and to experiment. Made a workflow specifically for it with things I use most in ComfyUI. Spritesheet maker was specifically for this, so made a version specifically for it lol. Rose in a flower pot. The basic image generator isn't fully connected since it's basic. Normally I'd add controlnets and ip adapters, but I figured I'd let you hook it up how you want lol.
Final Versions. I recommend the none Ollama one for reliable and consistent output, the Ollama version is more fun though. v4.2 and 4.0 have both seen updates to correct some issues and clean them up a bit. Things are more unified and Ollama has proper instructions now that work mostly lol
Updated after mistake: v4.2 to remove the forgotten prompt when I made the group from the nodes. But useful information, if you convert the text box to an input, the text stays lol. But 4.2 is corrected and I added a second Ollama node.
I MESSED UP lol, sorry guys, there's a hidden prompt I forgot to remove fox from. I'll fix this ASAP, I am so sorry I messed up with the grouping lol! I forgot to remove the prompt when I converted the text box into an input.
Update: v4.2 adds Ollama and IP adapter.
Version 4.0 - It may seem like some versions were skipped, they existed, I just forgot to share them as I got a bit hyper focused. 4.0 Needs an SD model, then it feeds the animateDiff video right into the spritesheet maker.
Text to spritesheet complete lol, now just needs fine tuning.
This doesn't actually require an SD model, I wasn't sure where to unselect that lol. My apologies for any confusion. Low tech solution in ComfyUI using some cool nodes in unorthodox ways lol.
version 1.1 ~ Finalized sizes. 40 sprites, 5 rows, 8 columns. 4096x2560 for super clean background removal, then shrunk down to 1024x640 to use as a sprite sheet where each sprite is 128x128. If you wish to use less than 40 frames you will have to adjust the math.
In my case for example, I wanted them to be 128x128 and I wanted 40 frames at 10 fps for super smooth sprite animation right? So, 8 columns x 128 = 1024 so my goal width is 1024. Number of columns (c) x sprite width (sw) = goal width (gw) so c x sw = gw
Now I have 5 rows, so 5 x 128 = 640. So my goal height is 640. Number of rows (r) x desired sprite height (sh) = goal height (gh) so r x sh = gh ~
Hello!
I'm your host, Geeky Ghost!
This is my sprite sheet maker simplified.
So, first up. Load image batch from directory. This feeds all the images in a folder to the Comic Panel Template node.
You can paste the same directory below it in the SameImageToLocal node. Or use a different directory so you don't accidentally send the batch back through with the sprite sheet lol.
Custom Panel Layout field
This one is important as well. H is Horizontal and V is Vertical.
H8 = 8 columns, having 5 8's means it will have 5 rows of 8, merged into a sprite sheet.
This is assuming it's a square image lol.
Remove Background
This one can take some tweaking. I'm trying to find a way to make it a bit easier. Some images may need some adjustment.
"Cleaned it up, it's more organized in my head :P when I try to explain it, the explanation tries to come out faster than I can type and etc lol. My apologies for the mess and mix ups.
Description
FAQ
Comments (24)
hi, how to install in stable diffusion? there is json in the archive
it's a workflow for ComfyUI, you just drag and drop the json file onto ComfyUI and it loads the nodes. Then you install the ones you'te missing by going to manager and clicking install missing nodes.
I had to change a few models because I couldn't find the models you specified. But then I get an error with the KSampler. Can you provide links to all the models you're using? Thanks.
nvm. upgraded torch to 12.4 and this is working. Thanks Geeky Ghost!
Is it right now limited to LCM models? In fact, it only works really well with your LCM model and not with many other models. Are there possibilities to expand this other than using post-processing?
I'm working on a flux version but need the other tools to catchup like controlnet and ip adapter and etc.
@TheGeekyGhost any news?
@johndeuff Unfortunately not yet, ended up getting sucked into some video projects that I may use instead. Added start and end frame to Frame pack and thinking that may be better based on the tests I've been doing with it. Still in my thoughts, but working on it in a round about way
How do I use this? I upload pictures but output says "text".
there's a couple of text inputs needed. There's two negative and two prompt boxes. I'll streamline it a bit and upload it.
This doesn't create the actual sprites, it only sorts them into a sheet? I'd be a little more clear about that.
It generates a video using the openpose images int, removes the background, then sorts the frames into a sprite sheet. So it creates them and sorts. Well, depending on which version you're referring to. some older versions only remove background and sort into sprite sheets, the rest create the video for the sprites animation, then remove the backgrounds, then sorts into sprite sheets.
An tip how to get / install the briarmbg background remover model?
Yeah, that would be really helpful. I found the safetensor file, but i don´t know, in which directory i need to put it, so the workflow will recognize it.
download the model from hugging face, the template uses briaai/RMBG-1.4. put the model in models/rembg (create the folder if not exists).
for the life of me I can't get this workflow to work though. KSampler keeps giving me an error:
'NoneType' object has no attribute 'shape'
@longemptyvoid777 I was getting that error but it seemed be because I was using the wrong a SDXL control net as I had it already. Getting the correct one worked for me.
@longemptyvoid777 work only briarmbg.pth, NOT safetensors.
I made a tutorial:
https://civitai.com/articles/24262
I need a video tutorial, because I'm kind of dumb.
Hi. I made, but not video ;)
https://civitai.com/articles/24262
Depends on NVIDIA/Windows, for all OSX users out there, can skip this one
anyone success? I want a tutorial
Hi. I made:
https://civitai.com/articles/24262
I made a tutorial:
https://civitai.com/articles/24262
Thanks man !

