General FAQ

What is Lumina?

Lumina is a free, open-source desktop application for translating manga, manhwa, and manhua. It automates the core pipeline — text detection, OCR, translation, inpainting, and typesetting — but keeps every step editable: full automation is rarely accurate enough, so a light human pass is expected to polish the final result.

Is my artwork uploaded anywhere?

No. All models run locally on your machine through ONNX Runtime, and page images never leave your computer. The only stage that calls an external API is translation, which sends just the text you are translating to the provider you chose (never the images). See the Privacy Policy.

Is Lumina open source?

Yes. The source code lives at github.com/lumina-tl/lumina and is released under the MIT License.

Model files are not covered by that license — they are downloaded separately and each carries its own license terms (for example, the default detector rtdetr is Apache-2.0, while the segmentation detector rfdetr_seg is restricted to academic, non-commercial research use).

Which platforms does Lumina support?

Currently Windows only. Lumina ships two Windows installers (DirectML and CUDA). macOS and Linux builds are not available yet — see the Installation FAQ.

Does Lumina need an internet connection?

Not for the core pipeline. Detection, OCR, inpainting, and typesetting run entirely offline once models are downloaded. Internet is only needed to download models and to use AI translation providers.

Does Lumina require an AI provider or API key?

No. You can run detection, OCR, inpainting, and typesetting without any API key. Translation is the only step that can call external AI APIs, and it is fully optional.

Where can I report bugs or request features?

Open an issue on the Lumina repository. For known issues and planned work, see TODO.md in the repository, and for the release history see the Changelog.