Automatic1111 Vs ComfyUI Test In detail | Stable Diffusion Automatic1111 and ComfyUI | Alfamix

so hey guys welcome back once again and in today's video what I'm gonna do is that we are gonna test both of the UI stable diffusion automatic one one one UI and a confi UI as well as you can see I have already set up both of the UI stable diffusion as well as config UI side by side so what I'm gonna do is that I'm gonna test different different sampling method and different different sampling steps as well in both of the uis and will show you the result the first thing and the very important thing that I will do is that I will not mess with different different checkpoint I am using Alpha mix over here thanks to one of my friend who has suggested me this model and you can see vae we are gonna go with VA as well and epic noise offset Lora as well so we are not gonna mess with all of these things we will also keep both of the prompts same we are only gonna mess with sampling method and sampling step as you can see this is already set up on automatic 1111 and you can see this is basically confi UI similar to this if you want me to explain let me explain you clearly so this is the chat checkpoint similar to what we select checkpoint over here on automatic 1111 and you can see a Lora loader over here this is positive prompt this is negative prompt and this here is the resolution I am using 7068 by 768 we are not gonna mess with that as well and the next thing you can see let me show you this is the sampler settings basically we are gonna mess only with the sampling methods and like steps as well so this is the thing and vae over here going on so this is basically the total UI that I have already created this is just similar to stable diffusion text to image as well so that's the thing that we are gonna do today now let's check up with these 20 steps as well and let's see what result do we get so guys that is the first result that we got with all our a sampling method and 20 sampling step so basically this is the result that we got pretty amazing result as well with automatic one one one now let me show you this result as well this is the result that I got as you can see this is uh 20 steps going on and all our answers style over here as well this is the result that we got pretty amazing result as well over here now let me show you one thing again so now as you can see on the left side this is the command window for comfy UI and on the right side this is the command window for table division automatic one one one so basically now you can see over here it took less than 6 seconds in order to generate this image with 20 steps is took basically less than 6 seconds and in this you can see it took like 2.45 its on stable diffusion and basically like I think so it took more than 6 second and eight like it's showing less than 8 second as well over here so that's the difference that I got with the 20 steps and now what I will do next time I will just close my stream live as well because I think so this is causing the speed also now what I will do is that we will set this to 30 steps let's go with 30 step here as well so now we will go with 30 as well and before generating I will close this as well because I think so this is causing this speed so let's close this as well and now let's check with the 30 sampling step and let's see what different does it make so guys now as you can see this is the result that I got after changing sampling save 230 definitely amazing result here as well and on confi UI as well you can see definitely amazing result here as well so basically this is also 30 steps one thing again I will show you this thing let me show you this time I got 3.20 its on confi UI on the left is confi UI and on the right you can see this is stable diffusion automatic one one one and I think so I got 2.7 its it took me like less than 11 seconds and this one it took me like less than 9 seconds so what do you think about this uh let me know in the comment section below I'm just making this video in order for you to understand how both of them are working I'm not against anything because I love both I'm user of an automatic one one one but just I'm making this for you in order to get some reference so nothing hate on any other I just love automatic one one one as well so let me know in the comment section below about this as well and now what we will do once again let's minimize this and now we will go with 40 let's go with 40 and on stable diffusion as well let's go with 40 step this time as well I'm gonna close my recorder uh OBS recorder and now with 40 step let's see what result do we get so guys as you can see this is the result that we got with the 40 step 40 sampling step and yes amazing result here as well we are not considering the result and this time on comfy UI this is the result that I got pretty amazing result right and you can see definitely 40 steps going here as well so now once again I will show you this is the result on the confi UI you can see 3.22 its and on this as well I got with the 40 steps you can see this I got 2.69 its over here so basically it took me like less than 50 seconds on stable diffusion and on confi it took me like less than 12 seconds over here so there is a difference definitely going on and you can see this result as well basically the earlier result 40 steps result was when I forgot to turn off my OBS and you can see while the OBS was on it got 2.86 its and on here as well let me show you so this is the result and in this I got like 2.66 its or 2.40 its and yes definitely OBS is making a difference uh speed difference as well so now what I'm gonna do we will try to change the model this time will not mess with the steps we will keep it 40 and this time what we will do we will go with different sampling method we will go with let's go with DPM plus plus 2m and here we will go with Kara's scheduler so we are going with DPM plus plus 2m Cara scheduler over here and on stable diffusion as well we will go with DPM plus plus 2m cares so basically we are going with the 40 steps DPM plus plus 2 m Karis and here as well so now let's see what result do we get let me close this first of all okay I will close this and and we will test with this method so hey guys this is the result that I got with the DPM plus 2 M Karas and 40 steps so this is the result that I got pretty detailed and pretty amazing result as well and if I show you the confi UI result so this is the result that I got with DPM plus plus now if I compare it with this speed see this now you can see this I got 3.26 its with 40 steps and less than 12 seconds and over here with automatic one one one I got 2.70 its and less than 15 seconds so basically that's the speed that I am getting and once again what I will do I will try to reduce we will go with directly 20 let's go with 20 over here as well let's go with 20 here as well and we will not mess with the sampling method and now let's check what result do we get this is the result that I got with the 20 steps you can see and DPM plus plus 2m cares so this is the result that I got pretty detailed result as well and on configure as well you can see uh steps 20 going on and sampler method DPM plus plus 2m cares so this is the result that I got and now if I show you there is result on command window let's go and let's check so you can see with the 20 steps it took less than 6 second and 3.26 its basically I am getting like 3.26 3.22 its compared to if you see on stable diffusion automatic 1111 with 20 step it took me like less than 7 seconds so basically it took like more than one second I think so so there's not a lot of difference but you can see the its yes definitely there is a speed difference and now once again what we will do we will go with like different sampler and I think so from my point of view DPM plus plus sde like is a heavy sampler so we will go with DPM plus plus sde and 20 step only here as well and here as well we will go with DPM plus plus sde caress so this is the sampling method we are testing with 20 steps now so let's check and let's try to generate first we will close the OBS as well so now guys as you can see after changing to DPM plus plus SD garage and and 27 this is the result that I got on automatic 1111 and definitely as I said earlier yes sde caress is definitely heavy I felt it from earlier as well and in on configure as well you can say DPM plus plus sde and scheduler carriers you can see this is the result that I got pretty amazing and yes pretty detailed result as well and if I compare it with this let's go with this so you can see I got 1.62 its this time a very low its and on automatic one one one you can see 1.33 its 420 sampling step as I said earlier yes definitely I feel that sde caress is very heavy sampling method and as you can see over here I got some warning error over here but I don't know what it is but still I got the result took around less than 12 seconds on confi UI and on automatic one one it took me less than 15 seconds so that's the thing that I got and once again this is the last thing we will do we will go with the 40 steps as well the last thing we will do with the 40 steps and we will go with 40 sampling step and let's see what result do we get and but first we will close this as well OBS so let's go so guys once again as you can see DPM plus plus sde carriers with a 40 sampling step this is the result that I pretty amazing result and pretty detailed result as well I just love this result but if I go with comfy UI as you can see yes that is the result that I got pretty detailed result over here as well but we are not considering with the result as you can see and if I show you this once again let's go on conf UI I got 1.58 its and yeah it took me like less than 25 seconds and on stable diffusion automatic one one one you can say I got 1.32 its and yeah less than 30 seconds so basically yes there is a speed difference going on over here as well and yes sde currency is definitely like a heavy carrier sampling method and yes guys that was the thing that I wanted to make so you you can get a rough idea about how both of the UI artwork basically both of the things are stable diffusion just think the UI is different this is automatic one one one and this is confi UI and the user interface this works on a node so I will keep on making video on confi UI as well on stable diffusion as well so guys if you like the video then do let me know in the comment section below what you want to know about each of the UI as well and if you want me to make another video such as like comparison or something like different things then do let me know in the comment section below and guys one more thing and thanks to Paulo I didn't know about this trick so if I want to load this UI once again you can go and copy this UI as well because I have already created it text to image UI checkpoints Lora loader positive negative from resolution and sampler setting vae and that's the thing that's the workflow that I was using you can go and copy this as well but if you want to load different UI just like this then there is a simple trick let me show you so basically if I go on to Output and if I just drag this on to this UI if I just drag onto this UI then basically it will load that UI you can see this is a different UI without a Lora without this is basically a different UI it's very simple to load UI you don't need to generate or create a UI again and again thanks to Paulo one of my friend recommended me this trick so it's very to load UI and once again if I show you this were the result let's drag this and let me drag this here as well and as you can see over here yes definitely the UI is reloaded here as well so it's very easy to load UI and yes definitely you can copy this UI as well and yes I am gonna make a different different video on comfy UI as well and just gonna go deep into this and that's the thing that I wanted to show you and yeah I will be back with more and more videos tomorrow till then have a good day bye guys love you all guys thank you for watching and yes bye guys

Channel: CHILDISH YT

Views: 16,124

Keywords: stable diffusion, stable diffusion tutorial, comfyui, comfyui guide, ai, ComfyUI Explained, stable, diffusion, Automatic1111 Comfy UI, Comfy Ui, Local ComfyUI, ComfyUI Installation, ComfyUI Local Install, Stable Diffusion Vs Comfyui, comfyui basics explained, comfyui nodes, nodes ComfyUI, comfyui tutorial, Automatic1111 Vs ComfyUI, automatic1111, Automatic1111 Vs ComfyUI Test, comfyui vs Automatic1111, ComfyUI Automatic1111, ComfyUI Test, ComfyUI Samplers

Id: DLzRd6jjAEA

Length: 11min 54sec (714 seconds)

Published: Thu Mar 16 2023