Apps that promise draw the face of your soulmate They do, in fact, deliver an image. It's usually beautiful, convincing, and specific. What it isn't: a prediction. It's worth understanding how this face is produced, because the process alone explains why it can't guess anything.
This text opens the box: what happens between you sending your photo and receiving the portrait, why the result looks so personal, and what the technology behind it actually does.
Where does the face that appears on the screen come from?
There are three techniques in use, either alone or in combination.
Geração por modelo de imagem
The most common type today. Generative models have learned, from enormous sets of photographs, how a human face is structured — proportions, lighting, skin texture, symmetry. From a command, they produce a new face that doesn't belong to anyone.
The essential point: the model generates a face. plausible, not a face existing. He was trained to produce images that look like photographs of real people. He doesn't have, and can't have, information about someone you haven't met yet.
Média de rostos
An older technique that is still used. The app combines features from several photos into an average face. There is a well-documented effect here: faces that result from averaging tend to be perceived as more attractive because averaging smooths out asymmetries and irregularities.
That's why the result is almost always pleasing. It's not about being in tune with your destiny—it's a statistical property of facial composition.
Transformação da sua própria foto
Some apps take the face you upload and apply changes: they alter gender, age, or specific features. The result bears a resemblance to you—which is often interpreted as a sign of affinity, when in reality it's the same source material.
Why does the result seem so certain?
Two things are acting simultaneously, and both are known phenomena.
The first one is familiarity. Faces generated from averages or from models trained on large sets of photos tend to look familiar because they share common features. Familiarity is often interpreted as recognition—the feeling of "I've seen this person before.".
The second is our predisposition to find correspondence. After seeing the portrait, it's natural that, upon meeting someone, the brain highlights what matches the stored image and ignores what doesn't. Because the portrait is generic enough, it will find similarities in many people—just as a vague prediction seems to be accurate.
What does the technology behind it actually do?
It's important to separate what is real from what is just a promise, because the technology is genuinely impressive in what it does.
- Generate images of faces that don't exist, with a high degree of realism. This works very well.
- Edit faces — Changing your apparent age, expression, hairstyle, style. That also works.
- Comparing two faces and to estimate if they belong to the same person. That's what facial recognition does, and it works within known limits.
- Predicting the appearance of someone you don't know. This doesn't exist because there's no input data. It's not a technical limitation to overcome: there's no information to start with.
The difference between the first three items and the fourth is the difference between processing existing information and creating non-existent information.
What to do with your photo before sending it.
Sending your face to an unknown app deserves two minutes of your attention:
- Check if the terms of use grant a broad license for the uploaded images — it's a common clause.
- Check if there is an option to delete the photo from the server after use.
- Be wary of apps that require phone registration for a purely visual function.
- Check the active subscriptions in the store after using it, because many activate recurring billing after a short trial.
A facial image is biometric data and deserves the same care as a password.
If the intention was to meet someone
That's where technology really helps, but in a much less mystical way: by putting you in contact with real people who are looking for the same thing. Dating apps don't predict anything, but they expand the circle of possible encounters, which is the factor that really matters.
A photo of yourself with your face uncovered, well-lit, and a profile that says what you like to do are worth more than any generated portrait. It's also worth taking the usual precautions: first meeting in a public place and having someone know where you are.
SEE MORE:
- Why can't any app predict a face?
- Compatibility in love: what can be measured?
- Safe apps for chatting with people nearby.
A simple test that debunks the promise in two minutes.
There's no need to discuss theory to verify what these apps do. Just run the same test twice, with identical data, and compare the images. If the result were a prediction about a specific person, it should always be the same. In practice, a different face appears each time, because the generation starts from a random point.
The second experiment is even more straightforward: ask a friend to enter your information into the same app on their phone. If the two results differ, there's no database query, just on-demand image production. If they happen to match, it means there's a small set of pre-existing images being randomly selected.
Here's a third test: enter an absurd birthdate, from a hundred years ago or next year. The app accepts, processes, and delivers a result with the same confidence, showing that the data serves only to fill in the narrative, not to calculate anything.
Why does the face come out looking pretty and generic?
Image generation models learn patterns from enormous collections of photos and illustrations. The result tends towards the center of this distribution: uniform skin, accentuated symmetry, studio lighting, neutral expression, and features that please almost everyone precisely because they are not striking. It is the average aesthetic of millions of images.
That's why the designs of different apps look so similar to each other. They all follow the same logic and are guided by similar descriptions, with requests for an attractive portrait and soft features. The face that appears on the screen isn't anyone's, and it's not exactly a random selection of real people: it's a statistical composite.
Details betray the origin when you look closely. Ears and teeth often appear irregular, the background appears blurred for no reason, strands of hair merge into the outline, and accessories like earrings appear different on each side. None of this indicates a defect in the application; it only indicates how the image was created.
Who owns the image that comes out of there?
This question arises when someone wants to use the design on social media, on a t-shirt, or on an invitation. The answer lies in the service's terms of use, and it varies considerably: some allow personal use only, some permit commercial use only in the paid version, and some reserve the right to reuse the generated design, including for promotional purposes.
If the image originated from a photo of yourself, there's an additional layer to it. Image rights are individual, and transforming someone else's photo into a drawing, publishing it, and associating it with a romantic theme can lead to embarrassment and civil liability. Only send your own photo, and ask for permission before using any other adult's photo.
Watermarks also deserve attention. Removing a service's watermark to publish the image as your own violates the accepted terms, and some services mark the file invisibly with metadata that identifies the source of its creation.
Step-by-step instructions for risk-free testing.
- 1. Download only from the official store and check the developer, date of the last update, and the data security section.
- 2. Register using a secondary email address and without logging in with your main social media account.
- 3. If you request a photo, use the limited selection of images instead of making your entire gallery available.
- 4. Run the test twice before considering paying anything and compare the results.
- 5. Decline notifications, which in this type of app exist to bring you back to the sales funnel.
- 6. Save the image if you like it, delete the account through the app itself, and uninstall.
Common mistakes are predictable: accepting the free trial just to see the version without a watermark, publishing the result with full name and date of birth visible, and submitting a photo of a third party for fun. All three create problems that last longer than a five-minute curiosity.
Frequently asked questions about app-generated drawings
O aplicativo usa a minha foto para treinar o sistema dele?
It depends on the contract you accept. Many services include a clause allowing the use of the submitted content to improve the product. If this possibility bothers you, the best course of action is not to send a photo: the final image does not depend on it for its production.
Existe versão que funciona sem internet?
Image generation requires heavy processing, done on a server. An application that delivers instant results without a connection is showing artwork already embedded in the package, with slight variations in color and framing.
E quando o desenho lembra alguém que eu conheço?
This is the expected effect of an average face. Common features remind many people of each other, and the brain fills in the rest by looking for similarities, which is the same tendency that makes you see figures in clouds. Two people looking at the same image usually remember different acquaintances.
Vale usar esse desenho como foto de perfil?
As an art form, it's a personal choice. In dating apps, however, a photo that isn't yours is a problem: profiles with generated images tend to be reported as fake, and the real-life meeting begins with an unnecessary explanation.
