Lumina

Appearance

Installation FAQ ​

Which installer should I download? ​

  • Lumina Setup — DirectML runs on any DirectX 12 GPU — including NVIDIA — and falls back to CPU on machines without a compatible GPU. Choose this if you don't have an NVIDIA RTX/GTX card.
  • Lumina Setup — CUDA is NVIDIA-only and highly recommended over the DirectML installer if you have an NVIDIA RTX/GTX card — it's noticeably faster. It requires a reasonably recent NVIDIA driver and is a larger download.

Both installers bundle the app and its Python backend, so no separate runtime installation is needed.

What are the system requirements? ​

  • 64-bit Windows (10 or newer is recommended)
  • A GPU is optional — CPU-only machines work, just slower
  • Free disk space for the app and any models you download (some models exceed 1 GB)
  • An internet connection for model downloads and AI translation

Do the installers include the AI models? ​

No. Models are downloaded on demand from inside the app (Settings → Models) or placed manually in the models directory. See Managing models.

How do I update Lumina? ​

Lumina checks for new releases on GitHub and can download and install them from inside the app. You can also grab the latest installer from the releases page at any time.

Is Lumina available for macOS or Linux? ​

Not yet. The current installers target Windows only. There is no published timeline for other platforms — follow the repository for announcements.

Can I run two Lumina installs side by side? ​

You can install the DirectML and CUDA variants separately if you want to compare them, but keep in mind they share the same app data directory unless you override it with LUMINA_MODEL_DIR for the models folder.

Do I need to install Python or ONNX Runtime myself? ​

No. The Python backend and ONNX Runtime are bundled with the installer and started automatically by the app. See the Technical FAQ if you want to run from a source checkout instead.