The tagger reads your location photos with OpenAI's vision models and proposes tags. Nothing is applied behind your back: you accept, edit, or throw out what comes back.
What it reads
From each photo it proposes:
- An architectural style, from Modern and Victorian through to Industrial
- An environment: interior, exterior, or studio
- A category: residential, commercial, industrial, or natural
- Features it can see in the frame, such as pools, gardens, stairs, and skylights
Where it is reliable, and where it is not
The tagger is dependable on things physically in the frame. A kitchen is a kitchen, exposed brick is exposed brick, a pool is a pool.
It is weaker on judgment calls. Architectural style is the main one: the line between mid-century and contemporary is a matter of opinion, and two scouts would disagree about it as readily as the model does. Treat style tags as a first pass to correct rather than an answer.
It also cannot see what is outside the photo. A house shot only from the front garden gets tagged from the front garden, so the tags are never more complete than your coverage.
Progress is shown per photo as a batch runs, and a photo that fails to analyse is skipped rather than taking the whole batch down.
Analysing a whole batch
Select several locations in your Locations view and choose Analyze Selected to run the tagger across all of them. Useful after a SmugMug import or a spreadsheet import, where thirty locations arrive untagged at once.
Before it starts, the dialog shows how many image reads the batch will cost and how many you have left this month. Trim the selection there if the number is higher than you want to spend.
What analysis costs
Each photo analysed is one image read from your monthly allowance. Uploading and storing photos costs nothing, so the spend is entirely under your control.
Three or four well chosen photos per location produce the tags you will actually search on. Running all forty photos of a house through the tagger costs forty reads and returns much the same set of tags. See Plans, Limits, and Image Reads for allowances and top-up packs.
Correcting what comes back
Accept the tags you agree with, drop the ones you do not, and add anything the photos could not show. Categories can be changed after the fact, the same as any other field on the location.