Stable Diffusion is a powerful AI model that can generate highly realistic images of faces from text prompts. When creating face prompts, it is important to provide sufficient context and detail so that the model can render the desired facial features accurately.
Some key elements to include in Stable Diffusion face prompts:
Subject and Framing
- Specify the subject (e.g. man, woman, child)
- Describe pose and framing (e.g. portrait, profile, full body)
- Mention age, ethnicity, or other identifying details
Facial Features
- Eyes (color, size, shape)
- Nose (size, shape)
- Lips (fullness, color)
- Hair (color, length, texture)
- Skin tone and quality
- Accessories like glasses or jewelry
Lighting and Background
- Lighting (hard, soft, rim, butterfly, cinematic)
- Background (solid color, bokeh, studio setting)
Style and Quality
- Realistic, painting, drawing
- Detail level (intricate, hyperdetailed)
- Output quality/resolution
Here is an example prompt using some of these elements:
Portrait photo of a beautiful young Black woman with warm brown skin and a radiant smile, facing forward with soft lighting. She has long, voluminous curls, bright hazel eyes with long lashes, and full lips. She is wearing a gold necklace and earrings. 4k resolution, extremely detailed.
Prompt Structure
Effective Stable Diffusion prompts tend to follow a similar structure:
Subject
Start by clearly stating the primary subject. For faces, specify details like gender, age, ethnicity if desired.
Examples:
- Portrait of an elderly Asian man
- Teenage girl with freckles
Description
Provide descriptive details of facial features like eyes, nose, lips, hair, and other identifying details relevant to your desired image. Be as specific as possible.
Examples:
- Bright green almond-shaped eyes, button nose, heart-shaped face
- Short messy red hair, thick eyebrows, 5 o’clock shadow
Framing and Posing
Indicate framing/cropping and posing to set the overall composition. Common options include:
- Portrait
- Profile
- Full body
- Facing left/right
- Looking up/down
Examples:
- Portrait, facing camera
- Full body profile, looking left
Lighting and Background
Add details about lighting, background, atmosphere to further enhance the image:
Examples:
- Soft studio lighting
- Cinematic rim lighting
- Out of focus lights in the background
Style, Quality, Details
Clarify art style if going for something painterly. Also indicate level of detail, resolution etc.
Examples:
- Photorealistic
- Watercolor painting
- Film noir
- 8K resolution
- Extremely detailed
Helpful Prompt Elements
Here are some additional prompt elements that can help enhance Stable Diffusion facial generation:
Subject Age
Clearly specifying an age helps render age-appropriate facial features:
Portrait of a 60 year old man with wrinkles around his eyes and mouth
Subject Weight
Adding weight descriptors results in appropriately fuller facial features:
Overweight woman with a round, chubby face
Ethnicity Details
Providing ethnic details helps generate facial structures and features typical of that ethnicity:
Indian woman with warm brown skin, almond-shaped eyes, and full lips
Emotion and Expression
Describe the desired facial expression to render appropriate emotion:
Young girl with a wide, joyful grin showing teeth
Accessories and Makeup
Mention glasses, jewelry, facial hair, cosmetics etc:
Woman wearing purple eye shadow and bright red lipstick
Lighting Direction
Specify lighting direction for consistent shadows/highlights:
Soft light from upper left, subtle catchlights in eyes
Image Quality
Indicate desired level of detail and output resolution:
Extremely detailed depiction, 8K resolution
Common Mistakes
Here are some common mistakes to avoid when generating faces with Stable Diffusion:
- Using overly generic descriptors like “beautiful woman”
- Forgetting to specify key details like hair color, age, lighting
- Providing contradictory details in the prompt
- Making prompts overly long and complex
- Expecting too much detail without indicating high resolution
- Not using enough negative prompts to filter out unwanted elements
Conclusion
The key to creating good Stable Diffusion facial prompts is to find the right balance of providing sufficient detail without overspecifying or making prompts overly complex. Start with the core subject and build up descriptive elements around facial features, framing, lighting and style to guide the model. Take an iterative approach, reviewing outputs and refining prompts until you achieve quality, consistent results.