MatchID finds the athlete.CaptionID names them.

In both GalleryID products. In GalleryID Desktop it runs locally on your Mac or PC against headshots you index there, so nothing is uploaded and it is unmetered because your own machine does the work. In GalleryID Cloud it runs against the gallery you uploaded, unlimited on all plans. Both write the same names to the same metadata fields.

For MatchID, yes. It matches faces against the headshots on your roster. But you do not need MatchID to caption: CaptionID writes the athlete name and the full caption line from a roster in one press or a shortcode, with no headshots at all. MatchID is the accelerator on top of that, not the way in.

The recognition built into consumer photo apps struggles here, because those models learned from photos where people face the camera. Action photography is the opposite: motion blur, heads turned away, athletes at distance, helmets. MatchID was trained on a large archive of sports photography instead, and reads jersey numbers when there is no face to read. It will not name everything on the first pass, and anything it is unsure of waits in a review queue rather than being guessed at.

It depends on the photos, and we would rather say that than quote a number that does not survive contact with your take. Face-visible sports like volleyball and basketball are the easy case. Helmet sports depend on how often a jersey number is legible. What we can promise is the shape of the failure: you set the confidence threshold, anything under it waits in the review queue, and nothing is written on a guess.

Yes, and that is what jersey number reading is for. Football, lacrosse, hockey, a catcher in the dirt: the face is hidden and the number is not. Numbers are read off jerseys, helmet decals, chest plates and shoulder patches, then matched to the roster. A helmeted face on its own scores lower than a clean one, so the number carries the frames the face cannot.

You catch it in review. The queue shows pending matches with their confidence score and the source crop, so you reject the bad ones and approve the good ones. The auto-approve threshold is your setting, which is what makes the claim "names are confirmed, never guessed" true on both paths: you confirm the name, or it clears a bar you set.

It is a step you trigger, not something running in the background, so it takes as long as the photos you point it at. In GalleryID Cloud photos process in parallel after upload. In GalleryID Desktop it runs on your own machine, and the usual pattern is to cull first and run MatchID only on the keepers, which is the fastest route to a named take. MatchID is never part of ingest: nothing analyses the pictures while a card is copying.

Facial recognition still works on partial occlusions. Sunglasses and masks reduce accuracy without breaking it, and jersey number reading is the backup when the face is not legible.

Any of them. Football, basketball, baseball, softball, soccer, lacrosse, hockey, volleyball, wrestling, track and field, swimming and diving, tennis, golf, cross country, cheer and rowing are the ones photographers run through it most. Identification is driven by your roster and your headshots rather than by a fixed list, so a sport we have never seen works the moment you load a roster.

No. Sports drove the engineering because it is the hard case, and the same engine works anywhere a known group of people gets photographed repeatedly: dance studios, school productions, music groups, conferences, corporate events, theater and portrait days.

None. Platforms that meter identification usually charge somewhere between one and three cents an image, which is cheap for one game and compounds across a season. MatchID is unlimited on all plans in Cloud and unmetered in Desktop, because in Desktop your own machine is doing the work. Per-seat pricing for this kind of software commonly runs $25 to $50 per user per month; GalleryID does not charge for users either.

No. There is no global index, no cross-client athlete database and no shared face library. Each account is its own MatchID context, your roster and headshots stay scoped to it, and we do not train models on your photography. MatchID was trained on our own photographic archive before any client uploaded a frame. In GalleryID Desktop the question does not arise at all: the photos, the headshots and the face data never leave your computer.