#Glossaries

Glossaries define preferred translations for names, terms, and repeated phrases, and CatScribe actively enforces them during translation. They are the single most effective tool for keeping a long project consistent.

#Where Glossaries Live

Open Glossaries in the sidebar. The workspace has five sub-tabs: Local, My Glossaries, Public Library, Favorites, and Attached. Local is the full desktop glossary manager described on this page; the other four are cloud-backed and show a connect prompt until cloud features are available.

#Groups And Entries

Glossary terms are organized into glossary groups. A group has a name, an optional description, and a language pair (for example en → pt). Groups are global per language pair — any job with a matching language pair can attach them; you select which groups apply per job at upload time.

Create groups by project, series, client, or subject area:

• One group for a novel series' character and place names.
• One group for a client's product terminology.
• One group for domain vocabulary (legal, medical, technical).

The in-app hint says it well: "Glossaries keep names and terms consistent across your whole book. Create a group first to organize your entries."

#Entry Fields

Under "Add New Entry Manually" each entry supports:

The entries list is searchable ("Source, target, or category...") and filterable by group, and paginates at 200 entries per page.

#Group Actions

Each group offers: "Set as active", "Stats", "Replace" (find and replace inside the group, with "Match case" and "Whole word only"), "Merge" ("Select groups to merge into this one. Matching terms are combined; the target keeps its own values on conflict."), and Delete. A global "Clear All" wipes every entry after a strong confirmation.

#CSV Import And Export

CSV is the exchange format (there is no TBX/TMX/XLSX support):

• "Import CSV" accepts multiple files. Column headers are auto-detected with fuzzy matching, the delimiter is auto-detected, files must be UTF-8, and language codes are normalized (pt-BR → pt). The result toast reports "Import completed: N successful, N failed, N duplicates (N total)".
• The expected columns: "Source Term, Target Term, Source Language, Target Language, Category, Group, Gender, Number" — for example hello,hola,en,es,Greetings,Default,,. Use "Export CSV template" to get a starter file.
• "Export CSV" writes your entries out — but note it is lossy: part of speech, the inflection flag, and variants are not included in the export. Keep the app's database as the source of truth for rich entries.

#Attaching Glossaries To A Job

In File Translation (with Auto Mode off), the glossary block is labeled "Glossary ({groups} groups, {terms} terms)". Select one or more groups whose language pair matches the job. Subtitle Translation has an equivalent multi-select.

Edge cases worth knowing:

• Language pairs must match exactly. The app warns: "No glossary groups for this language pair. Create groups in the Glossary tab, or switch Source/Target languages to match an existing glossary group (e.g. if your glossary is en → pt but you selected en → pt-BR here)."
• With no groups selected, file translation runs without a glossary ("No groups selected — glossary will not be used"). Subtitle translation behaves differently: with nothing selected it applies all glossary terms matching the language pair.
• Selecting multiple groups at once is a Pro feature ("Free plan: select 1 glossary group") — currently "Unlocked free during Beta".

#What To Add — And What Not To

Add decisions that must not drift:

• Character names, aliases, nicknames, honorifics.
• Locations, institutions, factions, organizations.
• Product names, technical/legal/domain terms.
• Repeated phrases with a fixed approved rendering.

Do not add every common word — placeholder protection means an over-broad glossary can make text stiff, and large term lists dilute the 12-term prompt budget for AI stages. Keep it focused on decisions that matter.

For a full sentence that needs exact wording in one place, edit the chunk directly or use a per-chunk re-run instead — see Chunk Overrides.

#Practical Workflow

1. Before the first run, import or add the obvious names and terms.
2. Attach the group(s) in File Translation and enable "Verify Glossary Terms".
3. Translate a short sample; check glossary QA flags in the CAT Editor.
4. During review, add newly discovered recurring decisions to the group.
5. Re-run only affected chunks (Chunk Override), not the whole job.
6. Keep groups small enough to maintain; merge duplicates with the Merge action.

For strategy across a whole book or series, see Maintaining Consistency and AI Translation Consistency.