Stable Diffusion Prompt to Fix Eyes

Generating realistic images of faces with stable diffusion can be challenging. One common issue is that the AI often struggles to render detailed, natural-looking eyes. However, with carefully crafted prompts and techniques, stable diffusion can create stunning portraits with lifelike eyes.

The key to improving eyes is using descriptive prompts that provide more context and detail about the eyes. Terms like “sharp focus,” “highly detailed,” and specifying the number, color, and gaze direction of the eyes can significantly enhance image quality.

In addition, leveraging other AI tools like GFPGAN for face restoration and Photoshop for manual touch-ups can further refine the end results. With practice and fine-tuning, artists can master directing stable diffusion to generate flawless eyes.

Prompt Structure

When prompting stable diffusion, treat the eyes as separate elements requiring unique descriptive parameters. Here is an example prompt structure:

A beautiful digital portrait of a woman with two large green eyes with thick dark lashes, soft skin, rosy cheeks, and red lips. The eyes are looking directly at the viewer with sharp focus and high levels of realistic detail.

Settings: Highres, sharp focus, intricate, highly detailed, digital painting, artstation headquarters, smooth, sharp eyes, upward perspective, centered

Breaking down the key components:

Eye Count

Always specify the number of eyes as “two eyes”. The AI can get confused by side profiles with only one visible eye.

Eye Color

Describe the iris color – green, blue, brown etc. More detail improves realism.

Eye Features

Include descriptors like “thick lashes”, “sparkling pupils”, “eyeliner” etc. The more the better.

Gaze Direction

Specify if the eyes should look straight ahead, left, right, up or down.

Eye Quality

Use terms like “sharp focus”, “highly detailed”, “intricate”, “realistic” to enhance clarity and definition.

Advanced Prompting

Take prompts a step further by providing the AI with more contextual references about the eyes:

A glamorous fashion model with piercing emerald green eyes accented by smoky eyeshadow. The eyes display extremely crisp focus and anatomy true to human eyes, including wet reflective corneas, visible iris textures, and defined limbal rings. The left eye gazes slightly upwards as strands of hair blow delicately across the model's face, which shows soft even skin with flushed cheeks.

Settings: Depth of field, intricate details, photorealistic, sharp focus, 4k, centered

Eye Makeup

Give more details like “smoky eyeshadow”, “winged eyeliner” etc. Spark inspiration from real makeup looks.

Eye Anatomy

Use anatomical terms like “corneas”, “irises”, “limbal rings” to guide medical accuracy. Reference eye diagrams if needed.

Gaze Nuance

Make eyes dynamic with cues like “gazes slightly left”, “eyes look downward”, etc.

Context and Storytelling

Set the scene, describe hair, face, and model type. This gives the AI more to work with.

Fixing Bad Eyes

If an initial image has issues with the eyes, there are several ways to improve quality:

Descriptive Rephrasing

Rewrite the prompt using more descriptive details about the eyes. Often new wording can enhance details.

GFPGAN Face Restoration

Run bad images through GFPGAN to resharpen facial features and reduce artifacts. This can correct malformed eyes.

Photoshop Retouching

Manually edit eyes in Photoshop using tools like the clone stamp to touch up details.

Inpainting Over Eyes

Use inpainting to completely replace both eyes, then generate new eyes based on an improved prompt.

With practice, stable diffusion can generate breathtaking eyes to match any creative vision. Use these prompting techniques to take control and consistently achieve photographic realism. Let descriptive prompts guide the AI and unleash its full potential for generating flawless eyes.

Useful Resources

  • Stable Diffusion WebUI:
  • Photopea (Free online Photoshop):