How to Take Photos for an AI Headshot Generator

Most disappointing AI headshots were decided before the generator ran. The face is slightly off, the smile belongs to someone else, or one eye sits a little higher than it should. People blame the model. Usually the problem is the eight photos that went in.

An AI headshot generator does not photograph you. It studies the pictures you upload, builds an understanding of your face, and then produces new images from that understanding. If the uploads are blurry, filtered, backlit, or all taken from the same angle, the model fills the gaps with guesses. Guesses are where the resemblance slips.

This guide is about the uploads. It applies to StudioGenie’s AI Photo Packs, which take 4 to 10 reference photos, and the same principles hold for most generators.

What the generator actually needs from your photos

It helps to think about what the model is trying to learn:

The structure of your face. Jawline, the distance between your eyes, the shape of your nose, your hairline. This needs sharp focus and even light.
How your face changes. A closed-mouth smile, a wide smile, a neutral expression. If you only upload one expression, every headshot will wear it.
How you look from slightly different angles. Straight on, a quarter turn left, a quarter turn right, chin a touch up or down.
What is consistent about you right now. Current hair length and color, current facial hair, whether you wear glasses.

Everything below follows from those four points. You are not trying to take a great photo. You are trying to give the model an honest, well-lit, varied description of your face.

Set up the light first

Light matters more than the camera, the outfit, or the background. Ten minutes spent on light does more for the final headshots than any other decision.

Stand facing a window during the day. Daylight from a window, with you facing it, gives soft light across the whole face. Turn off overhead lights so they do not add a yellow cast or shadows under your eyes.

Do not stand with the window behind you. Backlight turns your face into a dim silhouette with a bright halo. The generator learns a dark, low-detail face and produces dark, low-detail results.

Avoid harsh direct sun. Midday sun through a window creates hard shadows under the nose and brow. If the light is harsh, step back a few feet, or wait for a cloud. Overcast days are ideal.

Check for shadows across the face. One side of the face in deep shadow means the model only sees half of you clearly. Turn your body slightly toward the light until both cheeks are lit.

If you have to shoot at night, put a lamp with a plain white shade in front of you, slightly above eye level, and bounce it off a wall if you can. It will not match daylight, but it is far better than a ceiling light.

Use the rear camera, and get someone to hold it

The rear camera on a modern phone is noticeably sharper than the front-facing selfie camera. Selfie cameras also sit very close to your face, which stretches the nose and shrinks the ears. That distortion becomes part of what the model learns.

The simplest fix is to hand your phone to someone else and ask them to stand about five feet away, using the rear camera, with the lens at your eye level. If nobody is available, prop the phone against something at eye level, set a timer, and step back.

A few camera habits that help:

Wipe the lens with your shirt first. Phone lenses collect fingerprints and the resulting haze looks like soft focus.
Turn off portrait mode. The artificial background blur sometimes eats into hair and ears, and the generator learns those missing edges.
Do not zoom. Move closer instead.
Tap on your face to set focus before each shot.
Take three of every pose. Keep the sharpest.

Frame it like a headshot

The model needs to see your face clearly, so the face should fill a good portion of the frame. Head and shoulders is the right crop for most of the set. Two or three photos from the waist up are useful too, because they show how your shoulders sit and how you hold yourself.

Keep the camera at eye level. Shooting from above makes the eyes look larger and the chin smaller. Shooting from below does the reverse. A headshot generator will happily reproduce either distortion.

Look at the lens for most of the photos. A couple of shots with your eyes slightly off camera are fine, but the bulk of the set should be direct eye contact, because that is what a professional headshot needs.

Get variety, but keep the important things the same

This is the part most people get wrong. They either upload eight near-identical shots, or they upload photos from five different years.

Vary these:

Expression. A relaxed neutral face, a soft closed-mouth smile, and a genuine open smile. Think of something funny for the last one rather than holding a grin.
Angle. Straight on, then turn your shoulders a quarter turn each way and bring your eyes back to the camera.
Distance. Mostly head and shoulders, a few from the waist up.
Background. A plain wall, then somewhere with a little depth behind you. Nothing busy.

Keep these the same:

Your current look. Same haircut, same beard or lack of one, same glasses if you wear them daily. All photos taken within a week or two of each other.
Clothing that is simple. Solid colors, no fine patterns or large logos. Patterns can moiré and logos can be reproduced in odd places.
Natural skin. No beauty filters, no smoothing, no teeth whitening apps. A filter teaches the model a face that is not yours.

Recency is the quiet one. If you upload photos from three years ago alongside photos from last week, the model averages two versions of you, and the headshots look like a cousin.

A fifteen-minute shot list

If you want something to follow step by step, this set produces ten strong reference photos.

Face the window, plain wall behind you, head and shoulders, neutral expression, looking at the lens.
Same position, soft closed-mouth smile.
Same position, real smile. Have the person holding the phone say something ridiculous.
Turn shoulders a quarter turn to the left, eyes back to the lens, soft smile.
Turn shoulders a quarter turn to the right, eyes back to the lens, neutral.
Chin down a touch, eyes up to the lens, slight smile.
Step back so the frame includes your waist up, relaxed posture, hands loose.
Move to a spot with some depth behind you, such as a hallway or a room with the far wall several feet back. Head and shoulders, smile.
Same spot, quarter turn, neutral.
One more head and shoulders shot at the window with whichever expression felt most natural.

Review them on the phone at full size before uploading. Zoom in on the eyes. If the eyelashes are not sharp, retake that one.

What not to upload

Some photos actively hurt the result, even if they are lovely pictures.

Group photos with your face cropped out. The face is usually too small and often partly hidden.
Sunglasses, hats, masks, or hair over one eye. Anything covering part of the face has to be invented by the model.
Heavily edited or filtered images. Including old professional headshots that were retouched.
Screenshots, or photos downloaded from social media. Compression removes the fine detail the model needs.
Dramatic lighting. Neon, candlelight, colored stage lights, deep shadows. Great for Instagram, bad as reference.
Photos where you are laughing with your eyes closed. One is fine. Five is a problem.
Anything with another person’s face in frame. The model can borrow features.
Very old photos. If your appearance has changed, the photos should too.

None of these makes a pack impossible. They just lower the ceiling on how much the results can look like you.

Why the input matters more with AI than with a photographer

A photographer looks at you in real time, adjusts, and shoots hundreds of frames. If the light is wrong, they move you. An AI generator only ever sees what you give it, once, and then works from that.

That is also why the fix is cheap. You are not buying better software or writing cleverer prompts. You are spending fifteen minutes by a window. We wrote more about the mechanics of this in why AI photos stop looking like you, if you want the longer explanation of identity drift and how to avoid it.

How this fits a StudioGenie AI Photo Pack

StudioGenie packs ask for 4 to 10 photos. Ten is the maximum and a strong target, provided each photo adds variety rather than repetition. The pack then produces a curated photoshoot with at least 35 finished photos, all built from the same reference set.

Because every image in the pack draws on the same uploads, a good reference set pays off across the whole collection rather than in one lucky frame. The LinkedIn Headshots pack and the Executive Headshots pack both use the same upload step, so a single well-shot set works for either.

Once you have the photos, the next step is choosing a pack and reviewing the results for likeness, both covered in our guide to creating professional headshots with AI.

Common questions

How many photos should I upload? As many good ones as you can, up to the limit. For StudioGenie that is 10. Four sharp, varied photos beat ten repetitive ones, and ten sharp, varied photos are better than padding the set with duplicates.

Can I use my selfie camera if that is all I have? Yes, but hold the phone as far from your face as your arm allows, and keep it at eye level. The distortion is much smaller at arm’s length than at the usual selfie distance. Prop the phone up and use a timer if you can.

Do I need to wear business clothes in the reference photos? No. The generator handles the outfit. A plain, solid-colored top is all that matters, because the model learns your face and build, not your wardrobe.

Should I include a photo with glasses and one without? Pick one. Upload the version you want in the headshots. Mixing the two confuses the model about what your face looks like around the eyes.

Why did my AI headshots come out looking like a different person? Almost always one of three things: photos from different periods of your life, filters or retouching on the uploads, or a set where the face is small, dark, or partly hidden. Reshoot with the list above and the difference is usually obvious.

Can I reuse the same photos for a second pack? Yes. A good reference set is reusable. If you shoot it well once, you can reuse it for StudioGenie’s LinkedIn, Executive, Dating, or Lifestyle packs without shooting again.

Ready when your photos are

Shoot the set, check the eyes are sharp, and pick the ten best. Then choose an AI Photo Pack and upload them. The whole shoot takes less time than finding a photographer’s booking page.