Stable Diffusion is a highly sophisticated image creation AI that transforms text prompts into detailed, custom images. Given the complexity and specificity required by Stable Diffusion, creating the ideal prompt can be quite a challenge. This is where ChatGPT comes into play, aiding users in creating precise and effective prompts.
In this tutorial, we'll walk you through how to use ChatGPT to prompt for the Stable Diffusion 1.5 model.
Step 1: Understand Prompt Structure
It's crucial to understand the structure of a well-formed prompt for Stable Diffusion. Prompts consist of key elements or tags that guide the AI in creating the desired image.
An example prompt structure could be:
a portrait of a beautiful woman, topless, , (light freckles, beauty spots:1.2), she has long (dirty blonde:1.2) wavy hair, she is wearing a beige sunhat background of the ocean, sunset, sun-kissed, sunflare
Here, you can see how specific elements such as light freckles and dirty blonde hair have been assigned specific weights like :1.2 to control their prominence in the final image.
The use of parentheses ( ) and curly braces { } are important in assigning these weights. Parentheses denote standard weights, while curly braces are used to increase or decrease the weight of a tag. For example, {FUJIFLIM} would instruct the AI to render the image as if it were taken with a Fuji film camera.
Step 2: Start ChatGPT and Provide Your Request
Begin by starting a conversation with ChatGPT. Provide it with the details of the image you wish to create. The more detailed your description, the better the prompt it can generate. For instance:
Use this as an introduction to tell ChatGPT what to expect
You are a prompt AI for Stable Difussion AI, Stable Difussion is an image creation AI which is mainly used by receiving prompts and turning them into images, the only issue with stable difussion is its lack of consistency and difficulty to prompt demmanding users to use long and very technical prompts, this is where you come in handy, you will create the prompts for the user based on their request and make them be used in Stable Difussion.
🎨: I want an image of a dystopian cityscape at sunset, with towering skyscrapers, neon lights, and a silhouette of a lone wanderer.
Step 3: Wait for ChatGPT's Prompt
ChatGPT will then process your request and generate a detailed, technical prompt suitable for Stable Diffusion. This might look something like:
🎨: dystopian, cityscape, sunset, towering skyscrapers, neon lights, silhouette of lone wanderer, (deep shadows:1.2), (orange and pink sky:1.1), {wide-angle lens}, {low light:1.3}
Step 4: Input the Prompt into Stable Diffusion
Now, take the prompt generated by ChatGPT and input it into the Stable Diffusion AI. You can access Stable Diffusion 1.5 AI through the following link: Leonardo.ai
Step 5: Review and Refine
Review the image produced by Stable Diffusion. If it doesn't meet your expectations, you can refine the prompt with ChatGPT and try again. You might want to adjust the weightings of certain elements or add more specific tags to guide the AI.
Important Notes:
-
Always use technical tags in your prompts. Stable Diffusion is a machine, not a human, so make sure your instructions are as technical as possible.
-
Never create short prompts. The AI might crash if the prompt is too brief.
-
Always separate tags with a comma, or the AI might crash.
-
Remember the settings you used for successful prompts and adapt them for future requests.
-
Pay special attention to camera, lighting, and environment tags, as these greatly impact the final