Glossarion

Guides & docs

A quick tour of each feature and where to find it. Every section links to the matching chapter of the complete user guide, which explains every button and setting in plain language.

Novel translation

Choose an EPUB, TXT or PDF, pick a profile and model, set a chapter range and press Run. Context from recent chapters (or a rolling summary) goes with every request; long chapters are split into chunks; stopped runs resume from saved progress. Use the Progress Manager to retranslate chosen chapters and Refinement for a second polishing pass.

First translation → · Retranslation → · Refinement & output modes → · EPUB converter →

Glossary & NER

The glossary keeps names, genders and terms consistent. Pick a Glossary Mode to extract one automatically during translation, or press Extract Glossary to build it first and review it in the Glossary Manager. Auto-mapping attaches the right glossary to each book, and a unified glossary per language pair grows with every book you translate.

The optional NER name check uses a small named-entity model (a one-time ~109 MB download) to find names in the translated text that differ from the glossary, like "Gang" for "Kang", and fixes clear-cut cases. It is available in the NoCuda desktop build and in the mobile app.

The glossary system →

Manga & images

Open the Manga Translator (NoCuda desktop build, or Tools › Manga on mobile), add pages or a CBZ, and choose OCR, inpainting and font settings. The editor lets you run Detect, Clean, Recognize and Translate step by step and adjust boxes by hand.

Manga & comic translation → · Manga preview guide →

Vision, image editing & audio

The output mode decides what a job produces: Text, Vision (read text from images, then translate), Image (send pictures to an image-editing or generation model such as nan/gpt-image-2), Video, Audio (text-to-speech) or Refine. Image mode also works from a prompt alone, with no input file.

Output modes →

Local LLMs

Desktop: type ollamapull/<model> and Glossarion installs Ollama and downloads the model; use ollama/ or lmstudio/ for servers you already run, or route a custom prefix to any OpenAI-compatible endpoint. Mobile: open Model manager › On-device to download a Qwen3.5 preset or a custom GGUF and run it on the phone, or connect to Ollama or LM Studio on your PC.

Local AI and custom endpoints →

Multiple API keys & instances

Turn on the Multi API Key Manager to rotate many keys with automatic cooldown and failover, give pools their own jobs (fallback, glossary, QA, image editing…), and set a per-key endpoint or parameters. Give keys an Instance number to translate several files at once, each with its own keys and log.

Multiple API keys → · Multi-instance release notes →

Library & reader

Translated books appear in the Library with covers, In Progress and Completed tabs, search and filters. The reader has themes, fonts, page layouts, whole-book search and a raw/translated switch, and can translate the next chapter in place. The desktop library is in the standard and NoCuda builds; the mobile app adds a bilingual view and bookmarks.

QA scanner

Run QA Scan on a finished output folder, or let it run after every translation. It flags duplicate or missing chapters, leftover source text, AI artifacts, truncation, broken HTML, missing images and glossary name drift, and opens an HTML report you can work through chapter by chapter.

QA scanner →

Review generator

Generate an AI review of any EPUB in your target language — summary, characters, themes, style, strengths, weaknesses, originality scores and a rating — with spoiler and full-review modes.

Review generator →

SDLXLIFF editor

Translate .sdlxliff files and review them in an aligned source/target editor with a notepad view, machine-translation previews, a completed marker and score-based flags. The editor is in the standard and NoCuda desktop builds and in the mobile app.

Translation editing & review →

Subtitles, games & more

Previous← Getting started