ComfyUI Tutorial - Conditioning Timestep Range

all right uh I think I came up with the best example to talk about the conditioning time step range node and what it's actually doing so let's say you have two okay so steps so conditioning time step range it controls at what point during the step process to start using or to stop using certain conditioning values and the best way I can think of to explain that is if you think of steps in the most literal sense like you're literally stepping with your feet just imagine a staircase okay so you have you're telling a robot to walk forward up the staircase case and then you're telling it to also walk backwards up the staircase and if you combine those two things together it might be confused on what you actually want and it'll try to do both at once so walk forwards and walk backwards it might decide to walk sideways right and that might not be what you want because you might want it to walk forward part of the way and then at some other point start walking backwards and a visual example I can give you for that is actually this right here the example I was trying to use so we have two prompts one girl green here and one girl red hair so what I did up here is I we're ignoring this I combined them at the same time and what ends up happening the AI seems to get confused and it it kind of it it's basically doing a walk sideways like it's not doing green hair it's not doing red hair it's doing all things at once right and that's not really what we want in this situation so like let's say in this situation I wanted it to walk up the stairs for like a third of the way and then when it gets to that point it turns around and then starts walking backwards up the stairs after a certain point so that is what time step range allows us to do it allows us to stop using certain conditionings at certain points in the step process so that's what we're going to do here I'm going to replace the green here with well let's let's see what they look like individually at first we're going to generate one girl green here and see what we got one girl green here it's as expected and now we switch to the red here what do we get as expected okay but then you you when you combine them both at the same time you don't get green or red hair it's wait okay that's why I need to bypass this bypass so if you combine both of those prompts at the same time sorry for that mistake it it it's like that's that's the AI got confused right so time St range lets you specify at what point you want it to start doing something and stop so if we if we actually enable this now what's going on here let's follow the flow so we have one girl green hair with this this node and it says start with three so what does that mean it means so these are percentage values so 0o would be 0% one would be 100% so a value of3 means 30% so 30% is in reference to our steps so what this is doing is saying Hey I want you to start using this one girl green hair conditioning one once you get 30% of the way into the step process and then if you generate now see we're using this later in the stage so the first thing that it sees by itself is just one girl red hair and now we actually get a very interesting effect is even though we're using green hair conditioning in conjunction with the red hair conditioning after a certain point the color of her eye and the outfit actually changes as well because if we run the prompt just by itself with just the red hair red hair kind of Amber eyes I'm assuming they would turn red later with like upscaling or something and then her outfit's black so it's it's a very simplistic color palette but then just telling it to use you know a little bit more color information with that prompt later in the stage we get a really nice effect where it kind of keeps the rest of the image composition the same but we have more color diversity without it devolving into a you know this like whatever the heck this is supposed to be all colors at the same time uh so yeah that's that's pretty much how you use conditioning time step um but I I want to do one more example though I want to see what happens if we stop using this red hair so we're going to let's see so this starts so this one starts at0 three so I think we're going to see what happens if we stop using the red hair conditioning after a certain point so what's going on right here this is what's going on what we doing let's clean this up a little bit move these over have this right there okay so now we are telling the AI that from from zero so from the very beginning we're going to start using the red hair one girl but then we're going to stop halfway through okay so halfway through 10 steps we're going to stop and then 30% of the way in we're going to use the green hair conditioning and then we're going to continue using that up until the very end so we're just going to see what happens let's just see very there seems to be much stronger adherence to Green it's still very very similar to the red and that's because that's the first thing you start with but there is a much stronger Green presence now in the imagees you can see so that was oh what did I just do oh that's why okay so I just plugged this into the positive node and it didn't work that's because it so let's let's let's on I just made a mistake and I just kind of want to go over why it generates like this like why did it just come out Pure Noise so the reason is you're telling it to use conditioning for the first steps or whatever but then all of a sudden halfway through you're all like all right do whatever you want and it doesn't know what to do because it was never designed to just operate with with these these little fancy hacks that we like to use so that's what that's what just happened right there so what we're doing that that was a tangent we're just comparing comparing the original image with just the red hair was very strong red and black and white boring very boring and then with all of these things combined a much stronger Green presence better color diversity while still kind of adhering to the uh original look of what the image had so yep that's that's basically how this works um if you're curious about other things it can do I encourage you to experiment because that is literally how I learn all of these things so yeah uh if you have a question that is along the lines of hey what happens if you do this please try it because that is the only way anybody actually knows uh yeah goodbye

Channel: poisenbery

Views: 1,724

Keywords: comfyui, comfy ui, stablediffusion, ai, conditioning timestep, timestep range, conditioning timestep range, ConditioningSetTimestepRange

Id: MmPNJ4fowy4

Length: 6min 48sec (408 seconds)

Published: Thu Dec 07 2023