UX/People: Quick way to tag new faces based on similarity #2353

New issue

Open

opened 2026-02-20 01:10:16 -05:00 by deekerman · 0 comments

deekerman commented

2026-02-20 01:10:16 -05:00

Owner

Originally created by @causa-prima on GitHub (Jul 7, 2025).

Describe what problem this solves and why this would be valuable to many users

When I tag people, often there are many pictures of the same person. Currently, I can only tag people for a single picture, then move on to the next picture. This is very cumbersome and takes a long time, especially after the scan of many pictures (e.g. the initial scan of my library).

Describe the solution you'd like

For each tagged person, I imagine a list of recognized but untagged faces that are similar to that person, ordered by likelihood of being that person, with an easy way to tag that person (e.g. a ✅ next to that face).

This could be further improved by tagging whole clusters, i.e. faces that are regarded to be the same, but have no tag yet. However, that could lead to problems when faces within this cluster are actually not the same. But this again could be solved by using a two-step hierarchy, where you could have a checkbox on a cluster that would tag all faces within that cluster, with "sub-checkboxes" for each individual face.

As I described, I would see this as connected to a person. However, there could be a "clusters" subpage for the "New faces" part of the "People" page that lets you tag whole clusters similar to what I described for a person as well.

Describe alternatives you've considered

To me, this would be an improvement over simply allowing to tag multiple new faces as suggested in #1679 and is a bit similar to suggest similar faces for merging as suggested in #1595

Additional context
Very rough mockups, especially the one with cluster hierarchy would need to be improved to better visualize the "connection" within a cluster:
Without cluster hierarchy:

With cluster hierarchy:

Originally created by @causa-prima on GitHub (Jul 7, 2025). **Describe what problem this solves and why this would be valuable to many users** When I tag people, often there are many pictures of the same person. Currently, I can only tag people for a single picture, then move on to the next picture. This is very cumbersome and takes a long time, especially after the scan of many pictures (e.g. the initial scan of my library). **Describe the solution you'd like** For each tagged person, I imagine a list of recognized but untagged faces that are similar to that person, ordered by likelihood of being that person, with an easy way to tag that person (e.g. a ✅ next to that face). This could be further improved by tagging whole clusters, i.e. faces that are regarded to be the same, but have no tag yet. However, that could lead to problems when faces within this cluster are actually not the same. But this again could be solved by using a two-step hierarchy, where you could have a checkbox on a cluster that would tag all faces within that cluster, with "sub-checkboxes" for each individual face. As I described, I would see this as connected to a person. However, there could be a "clusters" subpage for the "New faces" part of the "People" page that lets you tag whole clusters similar to what I described for a person as well. **Describe alternatives you've considered** To me, this would be an improvement over simply allowing to tag multiple new faces as suggested in #1679 and is a bit similar to suggest similar faces for merging as suggested in #1595 **Additional context** Very rough mockups, especially the one with cluster hierarchy would need to be improved to better visualize the "connection" within a cluster: Without cluster hierarchy: ![Image](https://github.com/user-attachments/assets/8d62cb2c-bac4-4eb8-8fdd-ccad23a693b9) With cluster hierarchy: ![Image](https://github.com/user-attachments/assets/1e82be23-a3b1-40b9-af9b-6d3230a3c0f8)