If the AI keeps transcribing certain words in a way you don't want (for example, you prefer "甚麼" but the subtitles always come out as "什麼"), or you have proper nouns such as company names, people's names, brands, or products that often get transcribed incorrectly, you can use the Glossary feature to tell the AI your preferred spelling in advance. The system gives priority to these terms during transcription, cutting down the time you spend fixing them line by line afterwards.
When is the Glossary useful?
Fixed character forms or variant-character preferences: for example, always outputting "群眾" rather than "羣眾", or "甚麼" rather than "什麼".
Proper nouns that often get transcribed incorrectly: company names, people's names, brands, product names, technical terms, and so on.
Common particles or specific wording: when you want the AI to keep or standardize certain habitual words.
The difference between the "Workspace" glossary and "Custom" lists
The Glossary has two kinds of lists, each with a different purpose:
Workspace list: applied automatically to every project — good for shared terms like brand and product names.
Custom lists: not applied automatically; you choose which lists to attach when creating a project — good for terms specific to a field or an individual project.
In addition, the system automatically suggests terms based on your content; you can accept or skip them in the "System suggestions" tab.
How to add terms
Open "Glossary" from the left-side menu.
There are three ways to add terms — use whichever suits you:
Add one by one (quick add): type the correct term you want the AI to use directly into the table and press Enter to add it; you can also enter several terms at once, separated by commas. Each term can be marked as universal or set to apply only in a specific language, and you can add a short note to help the engine tell apart terms with similar meanings.
Import terms: bulk-import an existing term list from an Excel (.xlsx) or CSV file.
Extract from a document: upload a PDF, Word (.docx), TXT, or Markdown file, and the AI automatically extracts the terms in it for you to confirm and add.
Once added, the system gives priority to the terms in your Glossary in future transcriptions (workspace lists take effect immediately; custom lists must be selected and attached when you create a project).
Current limitations of the Glossary
Please keep the following two points in mind:
You can't set a fixed translation per target language.The Glossary can lock in how a term is written, but it currently does not support setting a fixed translated name per language for the same term. For example, you can't make "高川" always come out as "Takagawa" when translated into English. This improvement is planned but not yet available.
If you need this today, the current workaround is to finish transcription and translation first, then correct the translated names on each language's subtitle track line by line in the subtitle editor before exporting. You can also use "Find and Replace" to change identical translated names in bulk.
The Glossary is a guiding reference, not an absolute override. The AI gives priority to the terms you add, but it may still deviate in certain contexts, and we can't guarantee they'll be applied 100% of the time. If an individual subtitle still isn't right, you can fix it manually in the subtitle editor.
💡 Tips:
The Glossary is an existing feature available on all plans; the number of term lists you can create and the number of terms each list can hold vary by plan tier — the actual limits are as shown in the app.
Real-time transcription and translation (Live) also supports custom glossaries, so you can preload speaker names, proper nouns, and more before a live stream or event.