Post by Ka-Loc on Nov 1, 2022 17:45:59 GMT 1
Stable Diffusion, a popular AI art generator, requires text prompts to make an image. Sometimes it does an amazing job and generates exactly what you want with a vague prompt. Other times, you get suboptimal outputs. Here are some tips and tricks to get ideal results.
How to Write a Stable Diffusion Prompt
If you’ve spent any time at all with AI image generators, like Stable Diffusion, DALL-E, or MidJourney, you’ll have noticed that a well-worded prompt critical. A well-worded prompt is the difference between turning your idea into a great image and getting some uncanny-valley monstrosity with entirely too many fingers staring back at you.
Be As Specific as You Can
Stable Diffusion tends to thrive on specific prompts, especially when compared to something like MidJourney.
You need to tell it exactly what you want. Here is an example using the internet’s favorite animals: Cute cat.
Prompt: “Cute Cat”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
This applies to anything you want Stable Diffusion to produce, including landscapes.
Be descriptive, and as you try different combinations of keywords, keep a mental note of how the image changes. Some words and phrases tend to more strongly bias an image than other words, so you might need to adjust your prompt accordingly.
Name Specific Art Styles or Mediums
The need for specificity doesn’t end when you fully describe the content of the image you want. You can (and should) also specify the style you want. We’ll stick with our “Cute grey cat” prompt for this. Let’s say we want some cute grey cats, but we also want them to look like they’re an acrylic painting. The first thing you should try is adding “acrylic painting” as the next keyword in your prompt, so it would read: “Cute Grey Cat, acrylic painting.”
It is generally best to start with the fewest keywords to get close to what you want, and then add more to home in on an aesthetic you’re looking for.
Prompt: “Cute Grey Cat, acrylic painting”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
Stable diffusion recognizes dozens of different styles, everything from pencil drawings to clay models to 3d rendering from Unreal Engine.
Prompt: “Cute Grey Cat, Unreal Engine rendering”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
Those example prompts are extremely simple, but you can use dozens of keywords to fine-tune your results.
If you want weird, psychedelic-colored clay chicken models in a Walmart, you need to specify all those terms.
Prompt: “bust of a chicken in a walmart, clay model, (psychedelic colors), melting, weird” Sampler = PLMS, CFG = 7, Sampling Steps = 150
Name Specific Artists to Guide Stable Diffusion
Stable Diffusion is capable of doing more than emulating specific styles or mediums; it can even mimic specific artists if you want to do that. This example used Pablo Picasso.
You should definitely try throwing “by (Artist Name)” into your prompts if you haven’t already. It tends to produce dramatic results. Don’t be afraid to combine artists that wouldn’t normally go together, either. Oftentimes Stable Diffusion will combine their respective styles in exciting and unpredictable ways.
Good Luck
How to Write a Stable Diffusion Prompt
If you’ve spent any time at all with AI image generators, like Stable Diffusion, DALL-E, or MidJourney, you’ll have noticed that a well-worded prompt critical. A well-worded prompt is the difference between turning your idea into a great image and getting some uncanny-valley monstrosity with entirely too many fingers staring back at you.
Be As Specific as You Can
Stable Diffusion tends to thrive on specific prompts, especially when compared to something like MidJourney.
You need to tell it exactly what you want. Here is an example using the internet’s favorite animals: Cute cat.
Prompt: “Cute Cat”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
This applies to anything you want Stable Diffusion to produce, including landscapes.
Be descriptive, and as you try different combinations of keywords, keep a mental note of how the image changes. Some words and phrases tend to more strongly bias an image than other words, so you might need to adjust your prompt accordingly.
Name Specific Art Styles or Mediums
The need for specificity doesn’t end when you fully describe the content of the image you want. You can (and should) also specify the style you want. We’ll stick with our “Cute grey cat” prompt for this. Let’s say we want some cute grey cats, but we also want them to look like they’re an acrylic painting. The first thing you should try is adding “acrylic painting” as the next keyword in your prompt, so it would read: “Cute Grey Cat, acrylic painting.”
It is generally best to start with the fewest keywords to get close to what you want, and then add more to home in on an aesthetic you’re looking for.
Prompt: “Cute Grey Cat, acrylic painting”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
Stable diffusion recognizes dozens of different styles, everything from pencil drawings to clay models to 3d rendering from Unreal Engine.
Prompt: “Cute Grey Cat, Unreal Engine rendering”, Sampler = PLMS, CFG = 7, Sampling Steps = 50
Those example prompts are extremely simple, but you can use dozens of keywords to fine-tune your results.
If you want weird, psychedelic-colored clay chicken models in a Walmart, you need to specify all those terms.
Prompt: “bust of a chicken in a walmart, clay model, (psychedelic colors), melting, weird” Sampler = PLMS, CFG = 7, Sampling Steps = 150
Name Specific Artists to Guide Stable Diffusion
Stable Diffusion is capable of doing more than emulating specific styles or mediums; it can even mimic specific artists if you want to do that. This example used Pablo Picasso.
You should definitely try throwing “by (Artist Name)” into your prompts if you haven’t already. It tends to produce dramatic results. Don’t be afraid to combine artists that wouldn’t normally go together, either. Oftentimes Stable Diffusion will combine their respective styles in exciting and unpredictable ways.
Good Luck