How to Get a Great AI Headshot with Facial Paralysis

Gather Your Best Reference Photos

You need photos that show your face, not a diagnosis. The AI will copy what it sees. If you only give it flat, front-facing mug shots, that’s the stiff, symmetrical pose it will replicate—and that’s often exactly what you want to avoid. Think of this as giving the AI a vocabulary of your expressions.

Shoot 30-50 photos over two days. Capture different angles: your strong side, your affected side, a three-quarter turn. Get genuine smiles, thoughtful looks, even a quiet moment looking away from the camera. Use your phone. Natural window light is perfect. I once watched a client get her best shot while laughing at a text, her hand resting naturally on her cheek. That unplanned photo beat every posed one.

What makes a “best” photo? One where you recognize your own spirit. If your smile lifts higher on one side, that’s your smile. Show it off. Include a close-up and a medium shot. The AI needs detail in your eyes, skin, and hair.

Choose Your Lighting and Angle

Lighting is your sculpting tool. Harsh, direct light from a phone flash will deepen shadows in any asymmetry. Soft, diffused light from a window evens everything beautifully. Position the main light source—that window—slightly above your eye level and angled toward you. This creates gentle, defining shadows under your cheekbones and jawline.

Now, let’s talk about the camera angle. Should you face it straight on? Probably not. A slight turn of your head, about 30 degrees, toward your stronger side often creates a more dynamic and flattering line. It uses the natural plane of your face instead of fighting it. Test this. Take five shots facing the window directly. Then take five with your right shoulder dropped back slightly, then five with your left. You’ll see one angle where your features look most engaged.

Keep the light consistent across all your reference shots. Mixing shadowy indoor pics with bright outdoor ones confuses the AI. It starts guessing at your bone structure, and the guesses are usually wrong.

Upload and Tag Your Key Features

This is where most people rush. Don’t. When the platform asks you to tag photos, be brutally specific. Tag “smiling” photos only with genuine smiles where your eyes are crinkled. Tag “serious” photos where your face is truly at rest. Create a tag for your best angle, like “left side strong.”

Why does this matter? The AI blends these tagged categories. If you tag a tense, forced smile as “smiling,” your final headshots will look strained. I tell clients to sort their photos into three piles: “Yes, that’s me,” “Maybe,” and “No.” Only upload from the “Yes” pile. That pile might have 15 photos, and that’s enough. Quality drowns quantity.

What about the side with less movement? Tag it honestly. You might tag a photo as “relaxed right side.” This gives the AI clear data: this is how my face looks in this state. It won’t try to invent muscle movement that isn’t there.

Avoid the Common Symmetry Trap

Here’s the biggest mistake: asking the AI for a “symmetrical” face. It will try to mirror your strong side onto your weaker side, creating an uncanny, frozen look that resembles a wax figure more than you. Your face has character and history. The goal is a great portrait, not a geometry correction.

Instead, use the prompt box to guide expression, not anatomy. Write prompts like “warm, approachable smile,” “confident executive,” or “kind eyes.” Direct the mood, not the muscles. The AI is surprisingly good at translating emotional words into a cohesive expression using your actual features as a map.

One client insisted on “perfect symmetry.” The results made her cry—she said it looked like her estranged twin, not her. We redid it with the prompt “friendly and knowledgeable,” using her lopsided grin photos as key references. The new batch? She bought three. “That’s the me my friends see,” she said.

Review and Refine Your Final Images

Your first batch of 50 AI-generated headshots will be a mixed bag. Expect it. Your job is to be a sharp editor. Look for the images where your essence comes through, not where your face is most “even.” Does the expression reach the eyes? Is there life there?

Now, use the regeneration tools wisely. See a portrait you almost love, but the mouth looks stiff? Don’t start over. Use that image as a new reference for a second round. You can often note, “use this image but with a softer smile.” This iterative process is like tuning a radio—small adjustments get you to a clear signal.

Check for consistency. Does the hairline stay the same? Do earrings appear or vanish? These are tells of an AI struggling. Select the images where secondary details remain stable, proving the model has a solid grip on your core appearance.

Getting a great result is about control. You guide the process with precise references and clear, emotion-based prompts. The final image should make you think, “Yes, I look like that on a really good day.” That’s the win.

Related, and worth reading next: Train AI for Perfect Headshots With Your Service Dog.