Skip to content

Glossaries

Glossaries

A glossary gives the translation model preferred target translations for names, places, organizations, skills, and recurring expressions.

A glossary can also mark terms that should never be translated, such as brand names, as well as words that should never appear in the translation.

It is context for the model, not a guaranteed string-replacement engine. Important terms should still be reviewed in the resulting translation.

  1. Open Settings → Glossaries.
  2. Create a glossary and name it.
  3. Add source and target terms.
  4. Save each entry.

Global glossaries can be reused across comics. You can also jump straight to the glossary creation and editing page from Comic Settings.

In Comic Settings, use the Glossaries section to attach an existing glossary to the current comic. Multiple glossaries can be attached.

Avoid defining the same source term differently in multiple glossaries. If conflicting entries exist, the final result depends on their order and the model’s judgment.

In Comic Settings, enable Only add to prompt when matched in source text to include only terms that appear in the current page’s source text. When disabled, all glossaries attached to the comic are included. This can keep prompts shorter when you have many terms. It filters the context sent to the model; it does not replace source or translated text.

Use the import and export actions in Settings → Glossaries:

  • Import .gls.neko, .zip, .neko, CSV, and JSON files. If an archive is a full project package, you can extract its glossaries separately; an ordinary ZIP without glossaries has nothing to import. CSV/JSON imports use a field-mapping and preview flow. For CSV, choose whether the first row is a header; the delimiter is detected automatically. JSON accepts a top-level array or an array property in a top-level object.
  • A glossary with the same name (case-insensitive) is matched for merging by default; unmatched glossaries are created as new glossaries. Resolve each conflicting entry by choosing the incoming or local value before importing. Overwrite replaces the target glossary’s entries, including local-only or locally modified entries. Do not choose overwrite if you need to keep those entries.
  • CSV/JSON text is read as UTF-8. If the file uses another encoding, non-ASCII characters may appear garbled in the preview. Make sure the file is UTF-8 before importing.
  • Select one or more glossaries to export as a .gls.neko package. It contains glossary data, not comic images. This format transfers glossaries between Neko Translator installations and is different from a .neko project package, which contains comic-project data.

When exporting a comic project, you can also choose whether to include its attached glossaries. See Import and Export Projects.

Changing a glossary affects future requests. It does not rewrite existing source text or translations. Re-run translation for the boxes that should use the updated term, and choose the processing scope carefully.

Deleting a glossary or detaching it does not delete comic images, text blocks, or existing translations.