How to Get AI Headshots That Actually Look Like You

An AI headshot service turns a folder of ordinary phone photos into portraits that look like
they were taken in a studio. For a profile picture, a company page or a CV, the results are now
good enough that most people cannot tell — provided you feed it the right input. That is the part
worth getting right, and it is entirely under your control.

How do AI headshots work?

You upload a set of photos of yourself. The service learns what you look like from them, then
generates new images of you in settings you never posed for — a studio backdrop, an office, soft
lighting. You get back a batch, usually a few dozen, and pick the ones that work.

Nothing about the process is technical. The entire skill is in the upload.

What to upload, and what ruins it

This is the difference between portraits that look like you and portraits that look like your
cousin.

• Variety is the whole game. Different days, different backgrounds, different
  lighting, different clothes. Ten photos from one session teach the model one thing: that specific
  day.
• Only you in the frame. Crop other people out entirely rather than leaving
  them at the edge.
• Face clearly visible. No sunglasses, no hats pulled low, nothing across your
  face.
• Recent. Photos from five years ago produce a portrait of you from five years
  ago.
• Mixed distances. Some close, some from the chest up. All selfies at arm’s
  length gives every output the same slightly distorted perspective a front camera creates.
• Neutral expressions mostly. A few smiling is fine; twenty grinning photos
  produce a batch that all grin.

Heavily filtered photos are the most common cause of a poor result. The model learns the filter
along with your face and applies it to everything.

What to expect back

A batch, most of which you will not use. This surprises people, so it is worth stating plainly:
a hit rate of a few good images out of forty is normal. You are picking, not
receiving.

The recurring flaws are consistent across services and easy to spot once you know:

• Ears and jawline drift subtly on some outputs. Compare against a real photo
  rather than judging each in isolation.
• Glasses are handled inconsistently — frames warp or reflections behave oddly.
  If you wear them, upload plenty with them on.
• Skin gets over-smoothed, which is the main thing that reads as “AI portrait”
  to another person. Prefer the least retouched-looking option, not the most flattering.
• Hair at the edges can blur into the background on close inspection.

Before you upload your face anywhere

You are handing a company a biometric dataset of yourself, which deserves one minute of
checking. On the service’s own policy page, look for three things:

1. How long uploads are kept, and whether you can delete them on demand.
2. Whether your images train their models. Some let you opt out; some do not
   ask.
3. What rights you get to the outputs — particularly whether commercial use is
   included, which matters for a company page or a CV.

If a service does not answer those clearly on its own site, that is itself an answer.

Is it worth it against a real photographer?

Honestly: for a profile picture, yes. For anything where the photo is the product — a personal
brand, an author portrait, a company’s leadership page — a photographer still produces something
these do not, because a good one directs you rather than sampling you.

The gap is narrowest for a plain head-and-shoulders shot on a neutral background, which happens
to be exactly what most people need.

What a data policy has to answer

Uploading twenty photos of your face is not the same as uploading holiday snaps. Under the
GDPR’s Article 9,
“biometric data for the purpose of uniquely identifying a natural person” is a special category
whose processing is prohibited by default unless a specific exception applies. Similar
rules exist in several US states.

You do not need to read the legislation. You do need the service’s own policy to answer three
questions in plain language:

1. How long are my uploads kept, and can I delete them? A stated period and a
   delete control are the minimum. “As long as necessary” is not a period.
2. Are my photos used to train your models? Some services let you opt out. Some
   do not raise it, which is itself informative.
3. What rights do I have to the outputs? Personal use and commercial use are
   often different tiers, and a company page is commercial use.

If those answers are not findable in a couple of minutes on the service’s own site, that is
your answer about how carefully your face will be handled.

What it costs, and what that buys

Pricing in this category is per batch rather than per photo, and it moves often enough that a
figure quoted today is unreliable next month. The structure is stable:

• A one-off fee for a batch of generated portraits, usually a few dozen images
  from one upload set.
• More styles and higher resolution at the higher tiers, which matters if you
  need a print-quality file rather than a profile picture.
• Faster turnaround as a paid feature — free and cheap tiers queue.

Since you will discard most of a batch, judge value by how many usable images you get,
not by the count advertised.

Frequently asked questions

How many photos do I need to upload?

Most services ask for 10 to 20. Variety across them matters far more than the count — twenty
photos from one afternoon are worth less than ten across a year.

Do AI headshots look fake?

The bad ones do, and usually because the skin is over-smoothed rather than because the face is
wrong. Choose the most natural-looking output rather than the most polished.

Can I use an AI headshot on LinkedIn or a CV?

Yes — it is a picture of you. Check that your plan grants commercial use if it is going on a
company page, and use one that actually resembles you today.

Will they keep my photos?

It varies by service, which is why it is worth reading before uploading. Look for a stated
retention period, a delete option, and whether your images are used for training.

Why do some outputs not look like me?

Almost always the input. Photos that are too similar to each other, too old, or heavily
filtered give the model a narrow or distorted idea of your face.

Related: how to make AI images · the image generators compared