AI girlfriend from a photo: where the line is

You are on the character creation screen, there is a field asking for a reference image, and the face that comes to mind belongs to someone you actually know. An ex, a colleague, somebody from a group chat. Building an AI girlfriend from a photo is the one choice on that screen that can follow you out of the app, and it usually arrives at the exact moment you are deciding whether the paid tier is worth it, because appearance control is normally what the paid tier sells. So it is worth being plain about where the line sits and why.

What building an AI girlfriend from a photo actually does

An upload is not a face transplant. In most apps the reference image conditions an image generator, so the pictures the app produces lean towards that face, and a stored description or template of the appearance gets attached to the character. The text conversation is unaffected: the model has no idea who the person is, and nothing about the chat improves. What you are buying is consistency in generated pictures.

The second thing it does is more durable: the file goes to a company's servers and stays there under whatever the terms say. Deleting the character later does not reliably delete the upload, in the same way that cancelling a subscription does not delete a chat history.

This matters for the pricing question because photo upload, image generation and fine-grained appearance control sit on the paid side in most apps, while free tiers hand you a roster of preset characters and leave it there. The upload field is often the thing a free user can see but not use, which is exactly the kind of feature that gets you to upgrade on a feeling rather than on arithmetic. Our walkthrough of how to read a feature grid before you pay covers how to tell a capability from a bait.

Whose face it is

A photograph that lets someone be recognised is that person's personal data, which the UK regulator spells out in its guidance on what counts as an identifier. When you build an AI girlfriend from a photo of a friend, you hand a third party's data to a company whose terms that person never read, on an account they cannot access. You also cannot withdraw it for them.

Then there is the part with teeth. If the app goes on to generate intimate imagery of a real, identifiable adult who did not agree to it, that is no longer a question of manners. In the UK, intentionally creating a purported intimate image of another person without their consent is a criminal offence: section 138 of the Data (Use and Access) Act 2025 inserted it into the Sexual Offences Act 2003, and you can read the text of the provision yourself. Other jurisdictions have been moving the same way. The app's upload field is a form control, not permission, and the company on the other end is not the one holding the risk.

The rule that covers all of it is short. If you would not be comfortable telling the person whose face it is what you have done, do not do it, and no privacy setting changes that.

What the app's own rules tell you about the app

This is the genuinely useful part before you pay, because policies vary far more than the marketing does. Some companion apps publish a specific character creation policy: no photorealistic likenesses, no real-person references, and the restriction applied to thumbnails and in-chat images as well as the character portrait. Others carry a single line saying all characters are fictional and unrelated to real individuals. The first is a rule someone has to enforce. The second is a disclaimer that moves the consequences onto you.

So read the creator screen's own small print before the first charge, and look for four things:

• Whether real-person likenesses are allowed at all. If they are explicitly banned, uploading one risks the account you just paid for, and refunds for a terms breach are rare.
• What happens to the reference image. Kept indefinitely, deleted after processing, or retained for training. Apps that answer this in plain words tend to answer other questions in plain words too.
• Who can see it. Human moderation review is normal and reasonable, but it means a person may look at what you uploaded.
• Whether characters are private by default. Some platforms are partly social, and a character you made can be discoverable unless you say otherwise.

A vague answer on all four is not proof of bad faith, but it says the company has not done the work, which is fair to weigh against a monthly fee. The rest of that reading is in the pre-signup privacy checklist.

The etiquette, in five lines

None of this is about being polite to software. It is about the people who are not in the room.

• Presets and generated faces are the default. Most apps ship a roster, and the better creators let you describe a look in words instead, which gets specificity without involving anyone.
• Nobody you know. Not an ex, not a colleague, not someone from a dating app, not a public figure. Fame does not supply consent and neither does a public profile.
• Do not show it to the person it resembles. Telling them about a likeness does not make it a joke you are both in on.
• Check the frame. Your own photos usually contain other people, and a background is still a face.
• Your own face is your call, with one caveat. It is still a biometric-adjacent file on a company's server, and it is the single hardest thing to get deleted if the app changes hands.

What this means for whether you pay

If the photo upload is the main reason you are about to upgrade, it is the weakest reason on the page. It is the feature most likely to be restricted by policy, most likely to change without notice as rules on synthetic likenesses tighten, and the only one that can end the account outright.

The features worth a subscription are the ones that survive contact with a policy update: memory that actually holds across sessions, enough context that the character does not forget last week, voice minutes if you use them, and a flat fee rather than a currency you top up. Those are capabilities, and they are what the app we currently recommend charges for — on the usual condition that you set a reminder before the second month and decide again then.

The trade-off genuinely cuts the other way, and it is worth saying. An app with a fixed preset roster and no upload gives you less control over appearance than one that will take any reference you feed it. If a very particular look is the whole point of the thing for you, the restrained app is the worse product, and you are choosing between appearance control and a clean set of rules. What you should not do is let the upload field be the deciding feature while telling yourself it is a small one.

Questions people ask

Can I upload a photo of myself instead?

Generally yes, and the consent problem disappears. The storage question does not: your image sits on a company's server under its retention terms, so treat it as you would any other upload of your own face. Check whether the app lets you delete the reference separately from the character.

Will the app stop me if a photo is not allowed?

Sometimes. Apps increasingly screen uploads for real-person likenesses and photorealistic styles, and some reject them outright. Screening is inconsistent across the category, though, so an upload going through is not a judgement that it was fine — it may only mean the filter missed it, with the account suspension arriving later.

Is basing a character on a well-known figure any different?

Not in the ways that matter. A public figure has not consented either, likeness and publicity rights apply to them more actively than to most people, and app policies that ban real-person references do not carve out celebrities. The assumption that fame makes a face public property is not how any of the relevant rules are written.