In short: Confetti scans your connected photos, groups faces that look like the same person, and lets you put a name to that group once. From then on, every photo with that person is tagged automatically. It's opt-in, and Confetti stores the numeric face data, never a separate copy of your photos.
What does "face tagging" actually mean here?
A wedding shoot might have a few thousand photos and dozens of people in them. Face tagging is the process of figuring out which photos each person appears in, so you can filter, search, and build albums by person ("everyone with the couple," "just the bridal party") instead of scrolling the whole gallery looking for them.
How does Confetti find the same person across different photos?
Confetti runs a face-recognition pass over the photos in your connected folder. It groups faces that appear to belong to the same person into a cluster. You then name that cluster once, "that's Maria", and every photo in the cluster gets tagged with her name. New photos added later go through the same matching process.
Is face recognition "AI"?
Yes, this is one of the few places in the product that's genuinely AI-driven, as opposed to the album builder, which is mostly a rules engine (see our guide on Standard vs. Intelligent albums). Grouping faces by similarity is a machine-learning task; naming a cluster and getting it applied everywhere is where the real time savings shows up compared to tagging photos one by one.
What does Confetti store, and is it private?
Face grouping relies on computing a numeric description of each face (an "embedding"), not a photo of the face itself. Confetti stores those embeddings and the tags/names you assign, it does not take a separate copy of your photo library; your originals stay in your own Google Drive. Face data is processed only with your consent, and you can decline face tagging and still use the rest of the app. See our privacy policy for the full legal detail and our security page for the practical summary.
Can I fix a mistake if Confetti groups the wrong person?
Yes. Face grouping isn't perfect, especially in group shots, tricky lighting, or photos where someone is turned away from the camera. You can correct a mislabeled photo, split a cluster that merged two people, or merge two clusters that should be one person. Corrections improve that event's tagging; they don't require starting over.
Does this work for large weddings with big guest lists?
It's built for exactly that case: a big shoot with lots of people is where manually tagging photos one at a time stops being realistic. The bigger the photo count, the more time face tagging saves versus doing it by hand.