If I use all 8 fixes at once, won't that be more powerful?

No—it usually makes eyes worse. Each fix solves one specific problem. If you add all of them, you're giving 8 conflicting instructions and the model reverts to a default bland render. Use the diagnostic step to find your actual problem, apply one fix, and iterate. You'll spend fewer tokens and get better results.

Do these fixes work in all AI image generators, or just Midjourney?

They work in Midjourney, Stable Diffusion, and DALL-E 3, though phrasing varies slightly. The core instruction (sharp iris, positioned catchlight, direct gaze) translates. Start with the fix as written; if results don't improve, try 'sharp iris detail' or 'bright pupil reflection' as alternatives.

Why do professional photographers keep getting soft eyes while beginners sometimes get sharp ones?

Pros often over-specify (adding lighting modifiers, color grades, lens names) which can muddy the eye instruction. Beginners accidentally keep it simple. The fix: separate your eye instructions from your lighting instructions. Tell the model what eyes should look like, separately from how you're lighting the face.

Can I just use a higher resolution or upscale to fix blurry eyes?

No. Upscaling blurry eyes just makes them larger and blurrier. The fix has to happen in the prompt, before generation. Once the eyes are soft in the first render, resolution won't save you.

How many iterations does it usually take to get eyes right?

With the diagnostic method on this page, usually 1–2. Run the diagnostic, apply one fix, regenerate. If that doesn't solve it, apply a second fix. Most photographers lock in their signature eye render by iteration 3.

What to Type for AI

Why Your AI Headshots Have Blurry Eyes—And How to Fix It in One Prompt Edit

Name: Headshot Prompt Anatomy
Brand: What to Type for AI
SKU: WTTF-AZDKJAERSXKGFXXQVA7BTQ
Price: 29 USD
Availability: InStock

Blurry, unfocused, or lifeless eyes are the #1 reason portrait photographers reject AI-generated headshots. The problem isn't your generator—it's that your prompt is competing instructions that confuse the model about where to put detail.

This page walks you through the exact diagnostic phrases to identify which eye problem you're hitting, then gives you the word-for-word prompt edits that fix it. Most photographers solve this in one iteration once they know what to change.

Headshot Prompt Anatomy

$29

Pay once. Keep forever.

Stop burning 20 iterations guessing which word changed the lighting. Headshot Prompt Anatomy reverse-engineers exactly why professional AI headshots work—then gives you the reusable formulas to replicate those results for every client, in your own style....

One self-contained PDF. No hidden files or separate templates.

What's included

12 fully anatomy-broken-down headshot prompts—each with 6 labeled layers (subject, lighting, expression, camera, style, background) and 2–3 line edit notes explaining what changed and why
3 fill-in-the-blank master prompt templates: Corporate Neutral, LinkedIn Vibrant, and Creative/Editorial—each with a worked example showing a completed, client-ready prompt
Quick-reference card: the 6 micro-decisions that control headshot consistency (skin diffusion, eye catchlight, jaw angle, smile warmth, background blur, crop ratio)—each with a full spectrum of prompt phrases to choose from
5-iteration workflow system for going from first generation to locked signature style in 2–3 rounds
8 common AI headshot failures with exact diagnostic phrases and word-for-word fixes
Photographer-to-AI translation glossary: studio lighting setups, camera settings, retouching terms, posing directions, and color grades—all converted to prompt language you can paste immediately

Get it — $29 →

Or get free updates & new releases:

Follow for updates

The 8 Eye Problems—And What's Actually Causing Them

Every blurry or flat eye falls into one of these categories. Find yours, then jump to the fix section. **Soft/unfocused eyes (sharp face, mushy eyes)**: Your lighting or focus instruction is too vague. The model is rendering the face but not prioritizing eye detail. **Muddy catchlight (no spark)**: You mentioned light source but didn't specify where the reflection should land. Eyes read as dull without a bright, positioned catchlight. **Asymmetrical eyes (one sharp, one not)**: Conflicting instructions about head angle or camera position are making the model unsure how to render both eyes equally. **Closed or partially closed lids**: Your expression instruction is overriding eye clarity—'serene' or 'subtle' can push lids down. **Wrong eye color or missing iris detail**: You didn't specify color depth or iris texture. The model defaults to flat, washed-out rendering. **Crossed or misaligned gaze**: Your head angle and camera angle instructions are fighting. The model doesn't know if the person is looking at the camera or away. **Bloodshot or inflamed whites**: Over-sharpening or conflicting skin/eye instructions. The model is rendering texture in places that should be clean. **Tiny or recessed eyes**: Your camera distance or framing is pulling back too far, or your face emphasis is too weak.

The 8 Word-For-Word Fixes (Paste Directly Into Your Prompt)

Each fix targets one problem. Test one at a time so you know what worked. **For soft/unfocused eyes**: Add → `sharp focus on iris detail, crisp eye definition` **For muddy/missing catchlight**: Add → `bright catchlight centered in pupil, studio main light reflection` **For asymmetrical eyes**: Replace your head angle description with → `face square to camera, both eyes equally lit and in focus` **For closed/droopy lids**: Replace expression line with → `open, alert expression, eyes fully visible` **For flat or wrong iris color**: Add → `[specific color] iris, visible iris texture and depth, defined pupil` **For crossed or misaligned gaze**: Add → `direct eye contact with camera, gaze level and centered` **For bloodshot/inflamed whites**: Add → `clear, bright whites, no redness, clean sclera` **For tiny/recessed eyes**: Add → `medium close-up, eyes prominent and clearly visible in frame`

The One-Minute Diagnostic: Which Fix Do You Actually Need?

Before you start editing, run this quick check so you don't waste iterations. 1. **Look at the pupils**: Are they sharp and black, or soft and blurred? If blurred → use Fix #1. 2. **Look for light reflection**: Do you see a bright spot in the eye, or is it dark/matte? If matte → use Fix #2. 3. **Step back and compare eyes**: Is one noticeably different from the other? If yes → use Fix #3. 4. **Check the eyelids**: Can you see the full iris, or are lids covering it? If covered → use Fix #4. 5. **Check iris color**: Is it the color you specified, and can you see detail inside? If washed out → use Fix #5. 6. **Check where the person is looking**: Are they looking at the camera, or off to the side/down? If off-center → use Fix #6. 7. **Look at the whites of the eyes**: Are they clean white, or pink/red? If pink/red → use Fix #7. 8. **Check eye size in frame**: Are the eyes prominent, or tiny dots? If tiny → use Fix #8. Note the number(s), apply the matching fix(es) above, regenerate, and repeat.

Why These Fixes Work (And Why You Shouldn't Mix Them)

AI headshot models train on millions of professional portrait photos. But they also train on everything else on the internet. When your prompt is vague about eyes, the model doesn't know if it's rendering a closeup face, a far-away person, a painting, or a cartoon. Each fix works because it gives one specific, unambiguous instruction about eye rendering—borrowed directly from studio photography language the model already understands. The reason photographers mix fixes and still fail: competing instructions cancel each other out. If you say 'soft and dreamy expression' and 'sharp, crisp eyes,' the model has to choose which instruction matters more. Keep your eye fixes surgical. One problem, one fix, one iteration.

Quick Test: Lock Your Eye Settings Into a Master Prompt

Once you've found the fixes that work, add them to a template so you stop diagnosing the same problem twice. Take your best-eye headshot that worked. Copy the exact prompt that made it. Now highlight every phrase that touches eyes (catchlight, iris, gaze, lids, etc.). That's your eye baseline. For your next client, start with that baseline, then change only the subject and style variables. Keep the eye phrases locked. This is the difference between 'getting lucky once' and 'owning a consistent eye render across 50 headshots.' The Headshot Prompt Anatomy product includes a full micro-decision card for eyes—exactly which phrases produce which results—so you don't have to trial-and-error this for every client.

FAQ

If I use all 8 fixes at once, won't that be more powerful?: No—it usually makes eyes worse. Each fix solves one specific problem. If you add all of them, you're giving 8 conflicting instructions and the model reverts to a default bland render. Use the diagnostic step to find your actual problem, apply one fix, and iterate. You'll spend fewer tokens and get better results.
Do these fixes work in all AI image generators, or just Midjourney?: They work in Midjourney, Stable Diffusion, and DALL-E 3, though phrasing varies slightly. The core instruction (sharp iris, positioned catchlight, direct gaze) translates. Start with the fix as written; if results don't improve, try 'sharp iris detail' or 'bright pupil reflection' as alternatives.
Why do professional photographers keep getting soft eyes while beginners sometimes get sharp ones?: Pros often over-specify (adding lighting modifiers, color grades, lens names) which can muddy the eye instruction. Beginners accidentally keep it simple. The fix: separate your eye instructions from your lighting instructions. Tell the model what eyes should look like, separately from how you're lighting the face.
Can I just use a higher resolution or upscale to fix blurry eyes?: No. Upscaling blurry eyes just makes them larger and blurrier. The fix has to happen in the prompt, before generation. Once the eyes are soft in the first render, resolution won't save you.
How many iterations does it usually take to get eyes right?: With the diagnostic method on this page, usually 1–2. Run the diagnostic, apply one fix, regenerate. If that doesn't solve it, apply a second fix. Most photographers lock in their signature eye render by iteration 3.