[Feature] Write CLIP and OCR results to image metadata

[anaxonda]

I have searched the existing feature requests, both open and closed, to make sure this is not a duplicate request.

• Yes

The feature

Would like the ability to write the CLIP model results and OCR to image metadata, so that it is not just stored in immich database, and could be easily searched by external programs.

Could be added to xmp sidecar or direct to image exiftag (I'm not sure if immich ever writes directly to exif tags?)

This may be more practical for the OCR; for CLIP results a certain amount of keywords would have to be decided on I suppose?

Related to metadata handling here: #5530

Platform

• Server
• Web
• Mobile

[bo0tzz]

The CLIP data is just a numeric vector, it can't meaningfully be written to exif/xmp.