ComfyUI Fundamentals - Adv Masking and Compositing 

yeah, I started making tutorials because most tutorials ive seen just say, 'this is how you use this workflow' or explain things in a very brief overview sort of way. for example, using models and not explaining why that model or why that sampler or the relationships between various settings. everyone focused on just showing you how to use other workflows rather than explaining how to put things together and make them work. I try to teach people how to build workflows so they can use those skills to build more advanced workflows they might need rather than relying on workflows they dont fully understand and are unable to take full advantage of.

Just made a tutorial for external program integration if your interested :P still no live plugin for comfyUI but some day :P

happy to help :)

too long spent playing factorio to thank for that one to be honest :P

Still have to figure out timestamps but yeah, thats the plan, just hope people don't miss too much if they skip things.

i use timestamp to go back to unclear stuff if i need to, it's like having an index@@ferniclestix

started adding timestamps, gonna take a bit though, new ones will all drop with chapters though.

Impact Pack has a node which does this kind of thing, I am working on a tutorial covering impact pack but its kinda complicated.

@@ferniclestix thx a lot

one day lol, ive had intentions to do impact pack for a while, but you've seen how deep I go on things and theres so many options and things with impact its hard to do a comprehenisve video :P been working on one though.

pffft, you never know, i could secretly be an AI, you could be in the matrix right now and you would never know.....

eh, im home sick need to keep busy :P - as for things you can do with it, you could try compositing actual people into a background image and see what you can make :D pretty easy to make fake scenes from films this way.

huh, have not run into that issue before, I have however found that sometimes you have to use an image to RGB node if there are errors. no idea if that might help.

Thanks! @@ferniclestix

upgrade your comfyUI and restart browser.

I don't have a backend setup for this tutorial stuff yet so no discord atm sorry. and as for how well cut-off works, I spoke to the node creator and they basically said that its never going to be perfect. My advice would be to start using masking to force certain areas to be certain colors or using compositing to make your character in a seperate image and then bring it in, that will stop the color bleed. If you watched this tutorial you probably saw I said the character had a yellow biohazard suit, it made the room yellow, but when I paste him in another image, problem solved. If you combined cutoff and compositing you would probably have a whole bunch more control over colors.

thank you @@ferniclestix

huh... not sure which one of my vid's uses those specifically. latent switch was an impact pack node which they changed recently to 'switch (any)' and is now context sensitive. float to int is probably a quality of life pack node from memory and should still function as normal.

@@ferniclestix in part 4 video I installed through manager but when i restart comfyui it says import failed

do you know specifically which workflow is borked? like the name of the file because I loaded up my workflow from that tutorial and it loads and runs fine on my version of comfyui which is fully up to date.

@@ferniclestix wait i tagged you again in that workflow

yes, there are tools which allow you to do add or subtract from a mask, I believe impact pack has a couple of those too.

I note at the start of the video that using photoshop or gimp is probably a better move for most people, mainly this is to show nodes, techniques and teach skills that may be translated to comfyUI in general, not just how to copy past images into eachother in comfyui. You might like my how2 vids, I don;t have many but in them I explain stuff as quick as I can without explaining in depth.

Indeed, you are correct... from a certain point of view. photoshop is great for bespoke art related projects. it is not good for repetative automated batch tasks. The setup takes a while but once you have one creating images with this method is quite powerful. Although 100% if I wasnt doing automated generation or large batches of images i would use photoshop.

Of course, and this is one of the aspect the tool builders will have to innovate on.. The problem is - the tech is moving faster than the toolsets. If you really wanted to , you could simply integrate comfy with Krita (it already exists, twice over, maybe thrice over, i don't know). By fast , by the way, i mean : everything in the video above is now deprecated to IPAdapters, everyone uses some type of pipe instead of grouping 'buses', and comfy was updated 100 times minimum since. FLOSS can be chaotic. And that's a good thing :)

these tutorials are to teach skills and techniques using comfyUI, the purpose is to show how masking works and some of the ways it can be used to composite images together. I even point out right at the start that photoshop would be way faster. You may never need to use masks in this way, but then again, you may find yourself needing these skills when working with some specialized workflows, custom nodes or non standard work practices. by dave manes method, do you mean multi-area conditioning? havent covered that yet. but when used in conjunction with multi-area conditioning you can do some really really cool stuff with masking and compositing. :) all these skills tie together to allow people more freedom in interacting with comfyUI for multiple purposes. My goal is not to teach specific workflows or direct, do this and its perfect type tutorials, Im trying to teach general usage, skills and fundamentals behind the use of ComfyUI in this set of tutorials :) Great question though, hope I clarified things for you, and yes. ill be covering multi-area conditioning at some point, lotsa stuff still to do :)