The Glossary tab holds two culture-scoped tools: a term list that every translation job reads, and a free-text style guide. Both are per culture — switching the active culture shows that culture's own glossary and style guide. A project-wide brief for all cultures lives in Project Settings; see Project brief below.
Each row is one glossary term:
| Column | Meaning |
|---|---|
| Term | The source-language word or phrase to match (case-insensitive) |
| Translation | How to render it in the active culture |
| Do not translate | When checked, the term must be kept exactly as written — Translation is disabled for that row |
| Note | Free-text guidance for the translator model (for example, gender or grammatical case) |
Use Add term to append a blank row, the row's Remove button to delete it, and Save glossary to write your changes. A blank Term is silently dropped on save; saving refuses the whole list if any other required field is missing.
After a successful save, LocHub compares the glossary you just saved against AI drafts (status Draft, not yet approved or edited) whose source text uses a changed term but whose current translation does not follow it. For each such term, a banner appears:
"<term>" appears in N <culture> AI drafts that do not follow it. Apply to N strings
Clicking it rejects every one of those drafts with a note such as Glossary: translate "Sled" as "Nartas". (or, for a do-not-translate term, Glossary: keep "Sled" untranslated.). A rejected string is always resent to the model on the next translation job — it never reuses a cached answer — so the new glossary rule reaches it on the next run. Human-approved or human-edited translations are never touched by this: a human decision always wins.
<aside> 📝
Importing a CSV (below) triggers the same "Apply to N strings" check for whichever terms actually changed.
</aside>
A single free-text box per culture, saved with Save style guide. It travels to the model with every translation and repair request for that culture, alongside the glossary.
One text for the whole project, shared by every culture: what the game is, its setting and tone, who the player is, and anything else a translator should know before touching a single string. Set it in Project Settings > Plugins > LocHub > AI > Project Brief — there is no editor for it in the LocHub tab.
The brief goes to the translator and to the AI judge with every request, before the style guide and the glossary. A change applies to the running service automatically, the same way an AI Provider or model change does: right away, or — while a translation job is in progress — right after that job finishes. No restart step is needed.
Export CSV writes the current (saved) term list — blank rows excluded — as glossary-<culture>.csv, either through the editor's native save dialog (in the editor tab) or as a browser download. The file:
term,translation,dnt,note;\r\n line endings;