Remini: our review

Remini restores old, blurry or low-quality photos, with a stated specialty in handling faces — sharpening, feature reconstruction, colourisation. It's a consumer tool, distributed as a mobile app, published by Bending Spoons, an Italian company that also owns Evernote and Meetup. The central question for this tool isn't upscaling itself, but what happens to a face once it's sent to the app.

What it actually does

You upload an old family photo, a blurry portrait or a low-resolution capture, and the app reconstructs the features, increases resolution and removes digital noise. Beyond restoration, generative features let you create new images from a given face (avatars, staged scenes). It's this second category of features that changes things on the data side: the privacy policy explicitly distinguishes them from simple enhancement processing, with different retention rules for each.

Why it interests us

This is the most detailed case found in this category on the question of biometric data. The app's privacy policy precisely defines what it calls "face data" — the estimated position and shape of features — and acknowledges that this may, depending on applicable law, constitute biometric data in the strict sense. It goes further than a simple mention: training models on uploaded photos and videos requires explicit consent, off by default, and remains excluded outright for users based in the United States even if they turn it on — a geographic detail that betrays genuine legal thinking on the subject, likely tied to biometric laws in certain US states. For a firm or a company that has to explain to a client how a photo of their face is processed, this level of detail is rare and worth citing as an example, even coming from a competitor.

Where it falls short

Two reservations remain real. The first: the generative features send images and face data to named third-party providers — Black Forest Labs, Google and OpenAI — on top of cloud storage providers not detailed in the public documents. Using these features means accepting that the processed face circulates among several companies, not just Remini. The second: no pricing page exists. The model relies on in-app purchases, with free access funded by advertising, which prevents documenting a precise price before committing — a real gap for anyone who has to justify an expense upfront.

Our recommendation

For personal, occasional use, Remini documents better than its competitors what it does with a face, with model training genuinely optional rather than on by default. For professional use involving third parties' faces, the generative features should be turned on knowingly: they send the image circulating among several AI providers, not just the publisher.