Photoshop to Stable Diffusion (Single Node, updated)

[@jfosterdesigns]

Fantastic! Thanks so much for incorporating Nima's node. Hopefully, he'll incorporate your workflow and nodes into his plugin!

[@OriBengal]

Wow. Super helpful. Thanks.

[@CGFUN829]

Hey man , since u intersted in SD , have you took a look to tydifusion inside max ? its amazing and easy to use then nodes.

[@LinhLe-ib9gi]

i don''t connect PS with comfyUI . Image in Photoshop not show in ComfyUI, how to use it ?

[@albertrc92]

Hi! First of all thank you so much for your content. I've already used the workflow in Windows and it runs with no errors but now In a Macos it appears the error: When loading the graph, the following node types were not found: PhotoshopToComfyUI I've tried to install it via git and manually but it's still not working, any idea of why it's not installing correctly? Thank you in advance!

[@risunobushi_ai]

afaik most macOS users have been having issues with this particular node since my first video, and it's probably because of the way it's scripted. some people were suggesting downloading the latest release by Nima, that apparently might have re-introduced the PhotoshoptoComfyUI node back, but I haven't been able to check that yet.

[@tepa6816]

Hey Andrea, i like your content, specially the professional creatives side that sets some standards so people can have control for a change, rather than having AI spit out whatever it wants. However following this tutorial i am having issues with popup window not showing up. - I have also followed that suggestion you left in the other video(description) where you need to install requirements in the "comfyui-popup_preview" folder, but still its not giving a pop up window. - I am on windows 10 btw.

[@risunobushi_ai]

Hi! Thank you, can you try uninstalling it and installing it via git clone directly from the repo? github.com/NimaNzrii/comfyui-popup_preview There's a few users that have issues with the popup node but I can't understand what's the common factor among them

[@tepa6816]

@@risunobushi_ai I did it, also i run Stability Matrix enabled with admin rights but nothing changed. I also tried to start setup batch file in the comfyui-popup_preview/window folder to install what was required. -Then looked at a console where i've got this error: "E:/AI/Data/Assets/Python310/tcl/tcl8.6/init.tcl: version conflict for package "Tcl": have 8.6.9, need exactly 8.6.12 version conflict for package "Tcl": have 8.6.9, need exactly 8.6.12 while executing "package require -exact Tcl 8.6.12" (file "E:/AI/Data/Assets/Python310/tcl/tcl8.6/init.tcl" line 19) invoked from within "source E:/AI/Data/Assets/Python310/tcl/tcl8.6/init.tcl" ("uplevel" body line 1) invoked from within "uplevel #0 [list source $tclfile]"" Will see to update the environment//python package later when i get time. In the meantime, do tell if you have more suggestions. And cheers for the response!

[@risunobushi_ai]

Could be a dependency conflict, but if you installed it via git clone and manually requested the requirements from the txt file to be installed I don’t see why it shouldn’t have installed the proper versions. That’s very weird, but then again this is an issue some users have been having since my first video on this and I can’t figure it out.

[@tepa6816]

@@risunobushi_ai I have finally made it to work by ditching up Stability Matrix completely for clean Portable ComfyUI. So you might suggest this as well to others struggling with the same problem. Cheers again for the great content for the professional creatives! - Keep it up!

[@risunobushi_ai]

@@tepa6816 oh that's so weird! but then again there so many different user configurations out there that it's impossible to pinpoint weird issues most of the time. thanks, I'll add it to the description!

[@SejalDatta-l9u]

Recent subscriber here - great content! I keep getting this error. Do you have any ideas? "Error occurred when executing ImageScaleToMegapixels: NoneType' object has no attribute 'shape'" I've ensured that I've got all of the same models as you, although my picture in Photoshop is a different size. Thanks

[@risunobushi_ai]

mmh, it looks like the file that's being transfered from the Photoshop to ComfyUI node to the scale to Megapixels node is not an image. can you substitute the Photoshop to ComfyUI node with a Load Image node just for troubleshooting and check if you get the same error? This way we know if an image gets passed or not, regardless of the node that it's coming from.

[@SejalDatta-l9u]

​@@risunobushi_ai Great idea. Process of elimination. RE your point on PS not sending an image to Comfy. Is there a specific setting that I need to ensure on PS? I literally loaded an existing PNG onto PS kicked off the node with no additional steps.

[@risunobushi_ai]

@@SejalDatta-l9u did you open PS, go to edit - remote connections and opened the remote connection options, as well as setting the same password in the Photoshop to ComfyUI node?

[@SejalDatta-l9u]

​@@risunobushi_ai I really appreciate the quick repsonses. OK so problem solved. You were right. The issue was a comms break between photoshop and comfyui. Solution, follow Nima's (author of plugin) latest instructions using the CNX file. To use both your workflow and his new ones, I had to rename one of the folders under custom nodes since both use the same name, 'comfyui-photoshop'.

[@SejalDatta-l9u]

​@@risunobushi_ai I did everything that you did, exactly like your video, before I posted the comment. I was able to get it working by using Nim's file in his latest github.

[@saberkz]

Looks cool thank you for this video I have a request can you make similar but by drowing like anything in the picture then it get apply into SD preview ? For example a product photo then we drowing a sun and sky in canva and adding props etc then SD apply it on the frame

[@risunobushi_ai]

thanks! have you watched this one already? I covered some of those things here: kzbin.info/www/bejne/bXiUY4aXgdl4n6M

[@SejalDatta-l9u]

How do you keep the image produced by ComfyUI as similar to the images in Photoshop as possible. e.g. In the case of the female mode in your video, or even a landscape background, comfyui makes changes. While you might be able to solve the face using a face swapper, it might be a deal breaker for products or clothes that need to be exact. Any ideas?

[@risunobushi_ai]

by enabling the group that's under the main group! it's a remove background node with a blend image by mask node, that overlays the original subject to the newly generated image. it works best when used with the relighting group on as well, because otherwise the lighting and overall blending wouldn't make too much sense at times. otherwise you could lower the denoise a lot.

[@SejalDatta-l9u]

​@@risunobushi_ai Perfect timing! I was just playing around with those settings and found the same! Question on the Bland node: I found that it only processed the foreground when the mode was 'add'. When I switched another mode e.g. 'overlay' (doesn't work well), or 'color' (works well) I saw the background. I want to mimic the quality of your results. Any ideas?

[@risunobushi_ai]

ah, I didn't know Nima updated their repo, and I had recorded this video some time ago because right now I'm away from the office and I needed a evergreen video to publish. that's great though, it means we don't necessarily need to use the fork. for the Blend node, you want to use add because it works like a photoshop layer: with add at opacity 1, you're transferring the top layer on top of the bottom layer at 100% opacity, basically, where the mask is. So any other mode (overlay, darken, color) is not doing what needs to be done. the issue with transferring the original image is that the lighting is not the same between the top and the bottom layer, that's why it's better used with the relight group on - after the layer blend, the lighting is re-generated in order to blend the two layers taking into account the global lighting of the whole scene.

[@adberf]

Keep going Andrea, keep creating content. You are the best. 🤍

[@risunobushi_ai]

Thank you!

[@jccluaviz]

Seems popup its not working for me. It's not showing neither in Photoshop nor in any other window. What i'm doing wrong ? Many thanks.

[@risunobushi_ai]

Hi! What OS are you on? The popup node works for like 90% of users but there's some users out there that have some troubles with it and I can't seem to understand what's the common issue among those users

[@jccluaviz]

@@risunobushi_ai Thank you. I'm using Windows 10. Maybe something related with popup blockers or permissions? T'he popup its an Explorer Windows ?

[@risunobushi_ai]

It should be a window opened by python (like an executable sort of window) so there shouldn’t be any issues with pop up blockers. It’s just called pop up because it looks like a pop up, I think. I’ve got another user here who has troubles with it, I’ll let you know if they find a way to solve it.

[@agusdor1044]

Will this plugin works if I use comfy trough a cloud service like runcomy or runpod? Ty!!!!

[@risunobushi_ai]

unfortunately not, at least not in any way that I know of - you'd need to tunnel the remote ps connection and forward it to the cloud service, but that's a something that I'm not qualified enough (or a sysadmin enough) to know how to do, sorry :/

[@agusdor1044]

@@risunobushi_ai np my man, you are rockin in what you are doing. I will ask to support maybe, thank you!

[@coreypan8933]

helpful content,thx!!! But theres no a popup show up in photoshop, could anyone tell this ?😭

[@coreypan8933]

if ur pc is win,running this,it works. ...\ComfyUI-aki-v1.3\custom_nodes\comfyui-popup_preview\window\setup.bat

[@HanjLaoye]

The following node is missing: PhotoshopToComfyUI, how to solve it, thank you!

[@risunobushi_ai]

did you download it with git clone like in the video, following the provided link?

[@HanjLaoye]

@@risunobushi_ai yes，follow the video guide

[@risunobushi_ai]

Can you see the comfyUI-photoshop folder in the custom nodes directory?

[@DarioToledo]

Not sure that my 4gb of VRAM are gonna handle that 😅

[@risunobushi_ai]

I'm not sure about the minimum VRAM requirements (I really should start paying attention to that), but I think that the basic implementation (without relighting and subject masking) might be able to run on 4gb, give it a try!

[@DarioToledo]

@@risunobushi_ai I already use IC light (btw thanks for that extensive tutorial on product imaging), but I mean I'm not sure that 4gb will handle PS and comfy together with that many passages. Sometimes I tried having both up with an easy SD1.5 workflow and it went out fine, but with more conditioning (the depth node) might become a gamble.

[@dimitrishow_D]

I've been trying to make ai create my cartoon character in diff poses or with diff clothes...I must say everytime I try .. eventually I end up just doing it myself... It might be good for creating stuff out of nothing realistic persons or anime but it can't seem to work in my style since there is no real reference to it.. I tried training my own model but that also always ends up looking like a mentally handicapped person is trying to make my stuff.. So yeah it's fun but it's not good enough to create real specific stuff quickly

[@dimitrishow_D]

Or I'm doing it wrong lol

[@risunobushi_ai]

Hi! Sorry I forgot to reply to your previous comment. Training LoRAs is definitely not that easy, but I’ve seen some really good stuff coming out of LoRAs, even very specific ones. I’m not good at LoRA making at all, but there’s plenty of resources for that on KZbin and Reddit!