00:00hi guys welcome back with me Lord Chow Chow 2025 this time we're going to learn
00:05about how to run stable diffusion in using GTX 1660 Super that is my GPU
00:161660 Super is have a problem you just not setting it you know it usually gonna
00:23give you black screen or error screen so how to do it first from your stable
00:30diffusion installation select web UI and then select this web UI user bat open the
00:43file
00:47like this and you're gonna see common line arcs it's gonna be empty here
00:55so put this thing that I'm going to show you especially X warmer no half Y and no
01:04half this is the most important thing okay precision fuel upcast sampling the sub
01:13okay disable nunchuck you should I should use it so one X warmer no half Y and
01:23theorem okay and no half this is also important this is what gonna stop it of
01:29being error no half no halfway disable nunchuck X warmer gonna give you extra
01:35boost of power precision full and yeah the other I use it to increase my speed
01:44because you know someone say math your arm is not important for GTX 1660 but from
01:50what I'm experience is actually increase my performance so I use it anyway let's
01:57try it after that open the
02:02front man and it's gonna open up
02:12it's not gonna be fast of course but you know it were workable
02:21and I was I will show you how it generate image of course I'm going to skip this
02:29very long and boring thing so yeah I'm skipping until it's open okay I think we
02:40are done loading here you could see startup is 139 seconds after its launch we need to
02:48wait to load the model again it's depending on your model it could be very long it
02:53could be short but for regular it's like two minutes for GTX 1660 you know like you
03:01cannot load the idea SDXL you could load SD 1.5 and all its variants like I like
03:11photon like reality vision 1 through 5 maybe I use 4 and you know Zoya photo and other stuff
03:23this time I currently loading like the hello kit 25 I know that's the wrong model I use so it's
03:32load like that
03:35I need to repair that
03:39like that this is the model that you load you could use here I usually use wheat and
03:48head 768 you could use restore fees it doesn't matter but for now hires fix it
03:58takes a long time I don't know how to setting it up so never use it for now the
04:04other is working and you could also use LoRa if you have LoRa now let's make an
04:08example okay we're going to do this so let's generate okay I copy a G5 from it's
04:22hopping it's still not load because usually the first time you use it it cannot take
04:28a while load the first image after that it's going to be faster not crazy fast but just a
04:37little fast so it takes a while for the first image I usually just generate a lot
04:44of image if and then I just check the file or folder
04:51ETA also just don't believe it you know usually the last five seconds it takes a
04:58while but the real timing is here
05:04so it takes like nine seconds to do that okay it's gone but it's because of my
05:12it doesn't matter because the result is as available in the web UI outputs
05:19yeah I see two image and last thing here yeah this is the result
05:32it's not in the high-res fix so it's what you set is what you get this is the also
05:40the new image
05:43usually I need to make this work by this is the third image actually the first image
05:52usually as low and then later on is going to be fast not crazy fast like you expect because
06:02we use no half you know that's the problem no half and no half five is the problem okay now
06:10it's gonna
06:10work out we could also change like you can generate or you could select a seed that you want
06:22and make sure you do the right negative because this is really help you avoid any crazy
06:29man it's finally working like I want a snow winter and you know the moral things the dungeon like this
06:37so it's whatever and I said on a cell it should be in a cell whatever anyway guys I hope
06:45that helped
06:46thank you for watching and it's work you know you can use 1660 super per stable diffusion
06:56so don't worry if you have that and you know the time is not that time confusing even if you
07:02use Laura
07:03it's still not that bad
07:07especially use the model of SD 1.5 variants if it's based on SD 1.5 I think most of
07:17them is working I never
07:19found a model that are not working as long as based on SD 1.5
07:27SD XL you couldn't working because it needs more VRAM and the X1660 super cannot handle that amount of VRAM
07:39as much as I know at least anyway thank you for watching don't forget to like share and comment and
07:44see you again on the next episode have a nice day
07:48one
07:49one
07:49one
07:49one
07:49one
07:49two
Comments