About Skye Desk

Skye Desk is a macOS app for people who train LoRA models. It keeps the datasets, captions, training runs and results for each project in one place.

Why Skye Desk exists

Training a LoRA means juggling folders of images, hundreds of caption files, a cloud console, logs and sample grids. Skye Desk keeps all of that for one project in one window, so you always know which images, captions and results belong together.

How it works

Local first: projects are ordinary folders on your Mac, and the app never uploads them on its own.
Careful edits: caption changes are previewed before they apply, with history and undo.
Optional cloud: training on rented GPUs starts only when you ask for it. In the beta it runs in demo mode.
No account needed: sign-in will be optional and only for syncing preferences.

Where it is now

Skye Desk is in public beta. Dataset organization, caption tools and result comparison work today; cloud training runs on a simulated backend; accounts, sync and dataset health reports are on the way. The Download page lists known limitations.

Who makes it

Skye Desk is developed and operated by Skye Chiu. It is an independent product and is not affiliated with RunPod, Google, GitHub, Apple or Supabase.

Get in touch

Questions, bugs and ideas are welcome. Email us at support@skyedesk.app.