There's a ton prompt writing guides out there. This one won't be the best, but I hope it's not the worst.
My tenets for A.I. image creating are:
-There's no right or wrong way to write a prompt. A run-on sentence or a bunch of keywords separated by commas, either of those can work well. A run-on sentence equalizes every keyword, which means you need to ((push)) things that you need more of. A prompt with commas will push the earlier keywords, so they need to be rearranged to your preference.
-If you generated 100 images trying to make it perfect but hate all of them and need to stop, keep the 1 image you dislike the least, upscale and save it somewhere. There might be something about it that could help prompt writing later.
-Don't be afraid to increase or decrease the CFG if the prompt isn't working out.
The format for this guide will be to start with a basic prompt and improve it the best that I can. To add a bit of a challenge to this, I'll be using the original SDXL base checkpoint and an image dimension of 1344 x 768. There are far better checkpoints out and 1024 x 1024 is an easier ratio, but like I said, I want a challenge for this guide.
Check my profile for the ComfyUI workflow I use called "Everything Everywhere All At Once".
SDXL Base Checkpoint, DPM++ 2M Karras, 30 steps, 1344 x 768, 5 CFG, 253780601781563 seed
Positive: woman on a sidewalk
Negative:
I want the medium to be a photo.
Positive: photo of a woman on a sidewalk
Negative:
Getting close. Let's remove any other mediums.
Positive: photo of a woman on a sidewalk
Negative: 2d, 3d, cartoon, illustration, painting, sketch
Looking a little wonky at 5 CFG. Let's test some other CFGs.
3 CFG
4 CFG
6 CFG
7 CFG
I like it, but I wish she was younger.
Positive: photo of a 1girl on a sidewalk
Negative: 2d, 3d, cartoon, illustration, painting, sketch
That's way too young. "1girl" can be effective in a few scenarios, but I use it sparingly.
Positive: photo of a young woman on a sidewalk
Negative: 2d, 3d, cartoon, illustration, painting, sketch
I like it! Adding a bunch of stuff to it.
Positive: photo of a young woman walking on a city sidewalk at golden hour
Negative: 2d, 3d, cartoon, illustration, painting, sketch
"Golden hour" adds a really cool sunny glow, although it can dim the photo. Play with "sunny" "nighttime" "at night" "chiaroscuro" and "cinematic".
"Walking" can really loosen up a character's pose, although it can be too expressive.
I would like a full body shot.
Positive: full body photo of a young woman walking on a city sidewalk at golden hour
Negative: 2d, 3d, cartoon, illustration, painting, sketch
Some various camera zooms and angles you can play with later: full body, floor view, close up, extreme close up, high angle, from above, bird's eye view, low angle, from below, pov, side view, back view. It's best to put any of these at the front of a prompt for more weight.
Changing her ethnicity.
Positive: full body photo of a young Nigerian woman walking on a city sidewalk at golden hour
Negative: 2d, 3d, cartoon, illustration, painting, sketch
You can find examples of lots of ethnicities here.
An ethnicity can often change the clothing and setting of a prompt, so use wisely.
Changing her hair.
Positive: full body photo of a young Nigerian woman walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, 2d, 3d, cartoon, illustration, painting, sketch
Making her sporty.
Positive: full body photo of a young Nigerian woman wearing activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, 2d, 3d, cartoon, illustration, painting, sketch
Adding color.
Positive: full body photo of a young Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, 2d, 3d, cartoon, illustration, painting, sketch
"Colored activewear" leans blue. "Colorful activewear" adds a lot of colors, Specific colors, like pink or red, can bleed into other details and should be used sparingly, but SDXL is getting better at this.
Adding expression.
Positive: full body photo of a young cheerful Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body photo of a young smirking Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, 2d, 3d, cartoon, illustration, painting, sketch
"Smiling" can affect body language and zoom. "Cheerful" can affect body language. "Smirking" is subtle. If you have an angry face and not sure what to do, try this negative: "angry, unhappy"
I see a backpack in there.
Positive: full body photo of a young smirking Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, 2d, 3d, cartoon, illustration, painting, sketch
Adding detail.
Positive: full body highly detailed photo of a young smirking Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, 2d, 3d, cartoon, illustration, painting, sketch
"Highly detailed" is great for nudes but can add too many small details to clothing, so use it at your discretion.
Adding quality negatives.
Positive: full body highly detailed photo of a young smirking Nigerian woman wearing colorful activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
I have mixed feelings about all of those "worst quality" negatives, but some of them seem to help. I try not to overdo it. One negative I forgot to add to this, which I should've: "old, ugly"
I like how this has turned out so far, so I'll do a CFG check to find a sweet spot.
4 CFG
5 CFG
6 CFG
8 CFG
I rarely go higher than 6 CFG, so I'm surprised by 8. I think it's because I'm using the base checkpoint. Other ones seem to work better at lower CFGs. I'm changing to 8 CFG.
Adding a sexier outfit.
Positive: full body highly detailed photo of a young muscular smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Testing some body types.
Positive: full body highly detailed photo of a young muscular smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young smirking Nigerian female bodybuilder woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young curvy smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young obese smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Those five are the only ones I use. Most of the default characters in Stable Diffusion are lean and fit, so I don't need to use those keywords. But the above can really change the subject.
Combining two body types to test a full yet hourglass figure.
Positive: full body highly detailed photo of a young muscular chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Getting rid of the flab.
Positive: full body highly detailed photo of a young (muscular:1.1) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young (muscular:1.2) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young (muscular:1.3) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young (muscular:1.4) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Positive: full body highly detailed photo of a young (muscular:1.5) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
ComfyUI uses (muscular:1.3), which is the same as (((muscular))) in Automatic1111.
Going with 1.4, adding a darker complexion, and trying to fix her face a little.
Positive: full body highly detailed photo of a young (muscular:1.4) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: light skin, old, ugly, angry, unhappy, afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Running a CFG check to finish up.
5 CFG
6 CFG
7 CFG
9 CFG
Sticking with 8 and doing a final upscale.
SDXL Base Checkpoint, DPM++ 2M Karras, 30 steps, 1344 x 768, 8 CFG, 253780601781563 seed
Positive: full body highly detailed photo of a young (muscular:1.4) chubby smirking Nigerian woman wearing colorful form-fitting activewear walking on a city sidewalk at golden hour
Negative: light skin, old, ugly, angry, unhappy, afro, braided hair, curly hair, ponytail, short hair, backpack, blurry, grainy, low detail, low quality, worst quality, 2d, 3d, cartoon, illustration, painting, sketch
Coming next: Part 2 Advanced.