Stable Diffusion text2image is an Open Source AI image generator that uses deep learning to convert text prompts to images. Getting the best AI-generated image results depends on the prompts you input.
The quality of both the Prompt and Negative Prompt will directly affect the quality of your AI-generated artwork on stable diffusion. This guide will help you write prompts for stable diffusion.
How to Write Good Stable Diffusion txt2image Prompts
A Stable Diffusion prompt is a string of text that guides the model towards your desired image.
It comprises several key elements:
- Subject: The core concept you want the image to depict (e.g., “a cyberpunk city”).
- Attributes and Details: Qualities that refine the subject (e.g., “neon-lit, towering skyscrapers, flying vehicles”).
- Style and Composition: Artistic elements like camera angle, mood, and color palette (e.g., “low-angle shot, noir atmosphere, muted tones”).
Avoid abstract language. Use concrete nouns and precise descriptors to accurately convey your vision. Reference specific artists, movements, or techniques to guide the model towards a particular style. Refine your prompt gradually, testing different combinations of words and keywords.
Prompt weighting provides a way to emphasize or de-emphasize certain parts of a prompt, allowing for more control over the generated image. Prompt weighting also works with negative prompts.
In order to increase emphasis on a word or phrase, add a + or number between 1.1 and 2 at the end. A phrase can be including inside parenthesis. Example: (side basket of fries+).
In order to reduce emphasis on a word or phrase, add a – or number between 0 and 0.9 at the end.
The default weight = 1
Prompt Weighting Examples
Dinner table, plates, cheeseburger+, drinks.
Dinner table, plates, cheeseburger, drinks+++, people—.
Dinner table, (plates with cheeseburgers+++), drinks.
Mastering this skill requires practice, patience, and an understanding of Stable Diffusion’s capabilities. Here are a few prompt examples to help you.
Stable Diffusion Realistic People Examples
Prompt: Photo of young woman, highlight hair, sitting outside restaurant, wearing dress, rim lighting, studio lighting, looking at the camera, DSLR, ultra quality, sharp focus, tack sharp, DOF, film grain, Fujifilm XT3, crystal clear, highly detailed glossy eyes, high detailed skin, skin pores, 85mm lens.
Negative Prompt: Disfigured, ugly, bad, immature, cartoon, anime, 3D, painting, B&W, poorly drawn hands, bad composition, mutated body parts.
Stable Diffusion Checkpoint: epicrealism_naturalSinRC1VAE
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Prompt: Photorealistic black and white portrait photo of an old man crying, highly detailed glossy eyes, high detailed skin, skin pores, looking at camera, serious eyes, 50mm portrait photography, hard rim lighting.
Negative Prompt: color, crop, cropped, partial, disfigured, ugly, bad, immature, cartoon, anime, 3D, painting, bad composition
Stable Diffusion Checkpoint: epicrealism_naturalSinRC1VAE
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Stable Diffusion Architecture Examples
Prompt: Parisian luxurious interior penthouse bedroom, dark walls, wooden panels.
Negative Prompt: gold, pink
Stable Diffusion Checkpoint: architectureExterior_v40Interior
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Prompt: Architectural photorealistic cafe facade, modern and minimalist, weathered bricks, expansive glass windows, historical and contemporary architecture, soft natural light, clear afternoon sky, clean lines, minimalist design elements, ultra-realistic with many details, showcase patrons enjoying the ambiance.
Negative Prompt: Drawing, painting, B&W, cartoon, anime, bad composition, cars.
Stable Diffusion Checkpoint: architectureExterior_v40Exterior
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Stable Diffusion Product Examples
Prompt: sneakers in the style of Nike Airs, futuristic solarpunk, high detail, bright colours, clean composition, 3d render, standing on a pile of brightly coloured blocks, beautiful art work, intricate detail.
Negative Prompt: immature, cartoon, anime, 3D, painting, B&W, crop, cropped, partial, ugly, bad composition.
Stable Diffusion Checkpoint: epicrealism_naturalSinRC1VAE
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Stable Diffusion Animal Examples
Prompt: Photorealistic black and white portrait photo of black raven, dark background, highly detailed glossy eyes, ultra realistic, detailed feathers, bird, 50mm portrait photography, hard rim lighting, small details.
Negative Prompt: color, crop, cropped, partial, disfigured, ugly, bad, immature, cartoon, anime, 3D, painting, bad composition
Stable Diffusion Checkpoint: realisticVisionV60B1_v20Novae
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Prompt: Photorealistic black and white portrait photo of a wolf, serious eyes, highly detailed glossy eyes, ultra realistic, dark grey detailed fur, 50mm portrait photography, hard rim lighting, small details.
Negative Prompt: color, crop, cropped, partial, disfigured, ugly, bad, immature, cartoon, anime, 3D, painting, bad composition
Stable Diffusion Checkpoint: realisticVisionV60B1_v20Novae
Sampling method: Euler a
Sampling Steps: 50
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Upscayl to increase the resolution after image generation.
Stable Diffusion Vector Examples
Prompt: flat color, modern multi-color logo of a lion head, brown and yellow and black, on white background.
Negative Prompt: small, details, intricate, gradient, shadows, blurry, deformed, ugly.
Stable Diffusion Checkpoint: v1-5-pruned-emaonly
Sampling method: Euler a
Sampling Steps: 20
Width: 512
Height: 512
CFG Scale: 7
Seed: 736201470
Notes: Use Open Creative Cloud – Image to SVG Tool to convert the image to SVG. You can also use Vector Studio to generate SVG images directly in Stable Diffusion Automatic1111. View our Stable Diffusion Vector Image Guide for information on using Stable Diffusion to generate vector images.
Prompt: butterfly, flat color, multi-color coloring book drawing, bright colors, vibrant
Negative Prompt: small, details, intricate, gradient, shadows, blurry, deformed, ugly.
Stable Diffusion Checkpoint: coloringBook_coloringBook
Sampling method: DPM++ 2M Karras
Sampling Steps: 20
Width: 512
Height: 512
CFG Scale: 7
Seed: -1
Notes: Use Open Creative Cloud – Image to SVG Tool to convert the image to SVG. You can also use Vector Studio to generate SVG images directly in Stable Diffusion Automatic1111. View our Stable Diffusion Vector Image Guide for information on using Stable Diffusion to generate vector images.