Stable Diffusion Prompt Guide

[00:00] hey today I want to show you how I

[00:01] prompt in stable diffusion how I think

[00:03] about it when prompting and what methods

[00:05] I use to create prompts faster I'm using

[00:08] stable diffusion Forge UI and Juggernaut

[00:11] XL version 10 model but you can use any

[00:13] model you prefer with the settings that

[00:15] work best for that model many beginners

[00:18] simply prompt something simple like

[00:19] portrait of a woman while many short

[00:22] prompts still work in interfaces like

[00:24] Focus or AI like mid journey and chat

[00:26] GPT you're leaving too much freedom to

[00:29] the AI

[00:30] you can be more specific to get

[00:32] something closer to what you have in

[00:33] mind for example you can specify the

[00:35] type of image you want such as a photo

[00:37] illustration painting and so on as you

[00:40] can see even by mentioning photo the

[00:42] possibilities are still too many I got a

[00:44] vintage photo so maybe you can add an

[00:47] extra word like modern to specify that

[00:49] you want a modern photo instead of a

[00:52] vintage one to experiment better with

[00:54] the prompt you can use a fixed seed you

[00:57] can either use the last seed or just

[00:59] type number and generate again so we

[01:03] have the type of image and a subject but

[01:06] let's also place the subject in an

[01:08] environment maybe we can put her in the

[01:10] forest or on a beach or perhaps we want

[01:12] a studio photo with a black background

[01:14] behind maybe I don't want the woman to

[01:17] have a random hair color so I add blonde

[01:20] hair in the prompt perhaps I want her

[01:22] shirt to always be white so it stands

[01:24] out against the black background maybe I

[01:26] want the image to pop out more so I can

[01:29] add a rim light coming from behind

[01:32] visible in the hair alternatively I

[01:34] might want a golden hour type of light

[01:36] to make it look more natural if you have

[01:39] a specific hairstyle in mind you can add

[01:41] that to the prompt if you don't know the

[01:43] name you can search on Google for

[01:45] hairstyles women names for example if

[01:48] you like bangs you can add that to the

[01:51] prompt Additionally you can use chat GPT

[01:54] to provide lists for anything such as a

[01:56] list of women's clothing so I can use

[01:58] like a red cardigan instead of a white

[02:00] shirt let's try different types of art

[02:03] styles for the photo instead of just

[02:05] saying painting portrait which is a

[02:07] broad category we can specify further

[02:10] for example it could be an oil painting

[02:12] maybe with an old look or perhaps it's a

[02:15] watercolor painting we could also

[02:17] consider a pencil drawing but in this

[02:19] case we'd need to adapt the colors and

[02:21] background to match that style uh let's

[02:24] play around more with photography a

[02:27] cinematic fashion portrait could look

[02:29] cool

[02:30] we can also mention the nationality of

[02:33] the woman or the country for instance we

[02:36] could generate in African woman or a

[02:39] police woman since we've been so

[02:42] specific with clothes it might not

[02:44] include the police uniform so instead of

[02:46] a cardigan let's add a police suit or

[02:49] uniform let me quickly show you the

[02:52] negative prompt here you list things

[02:55] that you don't want to appear in your

[02:57] image it doesn't work perfectly some

[03:00] words are stronger than others but for

[03:02] example if you don't want a police badge

[03:05] you can add that to the negative prompt

[03:07] now when I generate again the badge is

[03:09] gone if you don't like the blue light

[03:11] behind you can add it to the negative

[03:13] prompt if that doesn't work you can try

[03:15] different seeds um in the worst case

[03:18] scenario you can use in painting to

[03:20] remove it let me remove the negative

[03:22] prompt and I'll show you how you can

[03:24] replace different words in the prompt

[03:25] scroll down and go to the XYZ plot on

[03:28] the X Type search for prompt Sr on the X

[03:31] values you put the words you want to

[03:33] search and replace first so for example

[03:36] I want to search for the word blonde and

[03:38] replace it with different colors like

[03:40] dark purple red and so on this way you

[03:43] can see how the girl looks with

[03:45] different hair colors as you can see now

[03:47] it started to generate the policewoman

[03:49] with different hair colors as a bonus at

[03:52] the end you have a grid with the

[03:53] searched word and all the other

[03:55] variation words that were replaced you

[03:57] can open it and see how they look in

[03:59] different

[04:00] colors let me dress her in a white shirt

[04:02] real quick let's say I want to redo

[04:04] everything but the girl's face to be

[04:06] more similar between Generations one

[04:09] thing that helps is to give her a name I

[04:11] usually use a random invented name but

[04:14] you can also use a mix of celebrities if

[04:16] you want so I'll give you a name and

[04:19] I'll generate again with that search and

[04:21] replace script active as you can see in

[04:24] the results the woman is quite similar

[04:26] between

[04:27] Generations because it has the same seed

[04:30] and the same description and name only

[04:33] the color is different let's say you

[04:35] like the results of a prompt but you

[04:37] just want a very subtle variation of it

[04:40] you can do that by adjusting the

[04:42] sampling steps or CFG scale you can do

[04:45] that automatically if you go back to the

[04:46] scripts and instead of prompt search and

[04:49] replace you choose CFG scale for the X

[04:52] values you can put between five and

[04:54] seven for example now when you generate

[04:57] you'll get those very subtle variations

[04:59] they might be harder to see on the

[05:00] screen but let's go to the folder where

[05:02] they were saved and switch from one to

[05:04] another as you can see some don't have

[05:06] badges or the light is a little

[05:08] different so you can choose the best one

[05:10] from all or you can combine them in

[05:12] Photoshop and keep what is best from

[05:15] each I also like to get variations for

[05:18] prompts using chat GPT so I have this

[05:20] prompt that I know works okay and I want

[05:23] a prompt adapted from different

[05:25] jobs I can ask GPT to adapt it for me

[05:29] and as you can see I get some nice

[05:32] variations let me test quickly the

[05:34] doctor prompt as you can see it added a

[05:37] white coat to the prompt now let's see

[05:39] how it looks like as a chef copy and

[05:41] paste The Prompt using the control plus

[05:43] C and Control Plus V shortcuts and

[05:46] generate quickly different variations

[05:49] sometimes I'm too lazy to write

[05:51] descriptive prompts so I just ask chat

[05:54] GPT to write it for me you just need to

[05:56] guide it a little in the right direction

[05:59] let's test it really quick to see what

[06:00] we get seems to work okay go to image to

[06:03] image and let me show you another trick

[06:06] that you can use when you don't know how

[06:07] to prompt but you have a photo or an

[06:09] illustration upload it here and then

[06:11] click on the paperclip icon that will

[06:14] interrogate clip which stands for

[06:16] contrastive language image pre-training

[06:18] the first time you're using it it will

[06:20] take some time to download a model you

[06:22] can see when it's ready in the command

[06:24] window but after that it should

[06:26] interrogate quicker that will give you a

[06:29] prompt for that image I'll copy this

[06:31] prompt and test it quickly the result is

[06:34] not bad could be better you can see that

[06:37] it's a policewoman but she doesn't look

[06:38] very similar to my image what seems to

[06:41] work better and get more consistent

[06:44] results for me is chat GPT I just upload

[06:47] a photo to chat GPT and ask it to

[06:49] describe it for me in a long sentence

[06:52] let's copy The Prompt and test it to see

[06:53] what we get now look at that how similar

[06:56] the photos look like with the image I

[06:58] gave it

[07:00] now you can talk with chat GPT to change

[07:02] the prompt how you want it for example

[07:04] you can adapt it for a doctor instead

[07:07] and this is the result if you don't have

[07:09] access to GPT or the internet and you

[07:12] still have problems with uh prompting

[07:14] but want to get some nice generation you

[07:17] can use art styles instead so if you

[07:20] have a short prompt you can go to art

[07:22] styles and either save your own art

[07:25] styles or download some free art styles

[07:27] from the internet I have an entire video

[07:29] with with 260 art styles and how to

[07:31] download them if you look at the prompt

[07:33] description then you look at the info

[07:35] for Generation you can see that you get

[07:38] the original prompt first then the art

[07:40] style is just another already saved

[07:42] prompt that is added to the original

[07:44] prompt plus the negative prompt let's

[07:47] talk a little bit about how to add more

[07:49] weight to certain words I have this

[07:52] prompt where I added green blue house

[07:54] next to each other to show you how I can

[07:56] make one word more important than the

[07:57] other so this is what I get for this

[08:00] prompt the house has blue and green in

[08:02] it I can make the words Blue House more

[08:04] important if I add round brackets you

[08:07] can see if you forgot to close a bracket

[08:10] it shows in red in the corner where the

[08:12] tokens are once I generate you can see

[08:15] the house has more blue than green you

[08:17] can add even more weight to those words

[08:19] if you add another

[08:21] bracket but there's an easier way to do

[08:24] this faster you just select the words

[08:27] then press control and use the up Arrow

[08:29] key to add more

[08:30] weight with control and the down arrow

[08:33] you can reduce the weight it can also

[08:36] have a negative value if you want it to

[08:37] be less important to change the value

[08:40] you select only the text without the

[08:42] brackets and numbers and use the

[08:44] shortcut again alternatively you can

[08:47] just change the numbers manually you

[08:50] should avoid going too high usually if

[08:52] you go more than 1.9 or two values it

[08:56] starts to go a little crazy and you get

[08:58] something like this like an a I exploded

[09:00] brain if I add a negative value to The

[09:02] Gnome for example it will be less

[09:05] important and you can see in the results

[09:07] that it's barely visible the main

[09:10] subject becomes the

[09:12] house here's how I usually prompt I put

[09:15] the art style or medium either first or

[09:17] last then comes the subject followed by

[09:19] the description then the environment and

[09:22] finally any extra information like

[09:23] colors lighting and mood while I was

[09:26] working on the video chat GPT released a

[09:29] new model version GPT 40 so now it's

[09:32] even easier to just ask for a stable

[09:35] diffusion prompt for what you want to

[09:36] create so if I want a watercolor

[09:38] painting of a cute bunny with glasses I

[09:40] can get a prompt quickly now I can test

[09:43] it in stable diffusion and the result is

[09:45] what I asked for I can then make more

[09:48] changes for example I can ask chat GPT

[09:51] to make it more minimalist and add a

[09:53] white background so now when I test it I

[09:55] get a nice cute cartoon bunny on a white

[09:58] background another thing you can do is

[10:00] to turn on generate Forever by right

[10:02] clicking on the generate button and

[10:04] choosing generate

[10:06] forever to stop it you need to

[10:08] rightclick again and choose cancel

[10:10] generate forever if you don't want to

[10:12] let it generate forever and have a

[10:14] specific number of generations in mind

[10:16] like between one and 100 you can move

[10:19] the batch slider to the value you want

[10:22] and it will generate that number of

[10:24] images alternatively if you have

[10:27] multiple different prompts you can

[10:29] choose prompts from a file or a text box

[10:32] you can paste prompts here or upload a

[10:34] text file with the prompts just make

[10:35] sure each prompt is on a separate line

[10:38] um I can once again use chat GPT to give

[10:40] me a list of different variations of the

[10:42] prompt with different animals then I can

[10:45] uh copy and paste that list of prompts

[10:47] in the text area when I hit generate it

[10:50] will start generating each of those

[10:52] prompts um as you can see I get

[10:53] different animals on a white background

[10:56] if you have any questions you can post

[10:57] them in my Facebook group pix Roma

[10:59] Community I usually post news prompts

[11:02] daily challenges design and crafts there

[11:05] I've recently passed 1,000 members so

[11:07] thank you to all who joined if you found

[11:09] something useful in this video leave a

[11:11] like thank you and have a great day