ClipWhat is a palette?
A palette (a "Log Selector profile") is your own metadata schema: the fields you log, each with a type, options, a plain-language definition, and optional key bindings, plus lookup tables and naming templates. One palette is live in both Logging and Browse, so a logged value is the same value you filter and rename by. Projects carry their own snapshot of a palette; your library keeps reusable templates.
What field types are available?
The palette schema supports Flag, Rating, Color, Text, Keywords, Single-select, Multi-select, Toggle, Number, Date, Subjects, and AI Description. Older profiles can also contain a Players field. Use Single-select for one mutually exclusive answer such as shot size, Multi-select for several attributes such as crowd and celebration, and Subjects for known people, items, or places. A rating expresses your assessment; an AI Description is a proposed narrative, not an identity decision.
Why is a palette worth setting up?
Three compounding reasons. It keeps your vocabulary consistent: a controlled list means "TD" and "touchdown" can't fork (you can even declare synonyms per option). It is the AI's output space: the model is handed your fields, options, and definitions and must answer in them, so better definitions produce better suggestions. And it drives naming. The palette carries both a display-name template and a written-file-name template, so your naming scheme is configured once and follows the project everywhere.
How do I set up a good palette?
Begin with the questions an editor will ask: who, what action, what shot size, where, and is it useful? Define a short controlled vocabulary for each select field. Add one free-text note for exceptions and a rating or flag for editorial judgment. Try the palette on ten varied clips before assigning it to a whole shoot; remove fields that do not help retrieval or handoff.
Can the AI invent values my palette doesn't have?
Each supported field has a setting controlling whether AI may propose new values. Keep controlled vocabularies closed when consistency matters, such as shot size or a delivery category. Allow new suggestions only where the field benefits from discovery. Review a proposed new value before using it to expand the team's shared vocabulary.
What are lookup tables and code-replace?
A lookup table is any imported code→value sheet parsed from CSV/TSV (paste or file): a jersey roster, location codes, shot codes, crew, SKUs. Bind it to a modifier layer and logging becomes type-the-code, read-the-name: hold the layer, type "7", and the HUD offers every match by code, name, or detail column. You can join multiple CSV columns into the stored value (first + last name), keep display-only columns out of it, and replace a table's contents in place later without breaking the bindings that point at it.
Can I share or move a palette between projects and machines?
Yes. Palettes export and import as .loupeprofile JSON files, and attaching one to a project copies it in (so project edits stay local). A project's palette snapshot travels inside the project document, over a synced relay or on the volume itself, so the schema and naming scheme arrive wherever the project does. Profile files are tolerant across versions: an older build reading a newer palette degrades unknown field kinds to text rather than failing.
How does the palette map to real metadata on export?
A field's type and target determine where its value belongs. Native fields such as rating and note differ from custom structured fields. Export formats support different subsets and representations, so test the fields that matter in the receiving application. Matching column names alone does not establish a complete round trip.