Which platforms are supported?
macOS supports Apple Silicon 11+ only; Windows supports 10 22H2 / 11 x64 only.
FAQ
A few decisions before downloading
Short answers and next steps live here; complete commands and setup steps stay on the docs page.
Can I run it with 8 GB VRAM?
Verified 8 GB Turing sm75 cards use the compact tier; Z-Image Turbo still requires 12 GB.
How much disk should I reserve?
Each generative model is about 25–33 GB; reserve roughly 40 GB per model and about 150 GB for the full set.
Is it completely offline?
Inference and work management do not use a hosted service; model preparation, update checks, and off-site links still access the network.
Why does the system warn about the installer?
Current artifacts are unsigned. Verify SHA-256, then follow the documented system-prompt steps.
How do updates work?
Get new versions from official GitHub Releases and verify the digest before installing; the site makes no unsupported auto-update promise.
Can a failed job be recovered?
Jobs, attempts, and formal works are managed separately; failures can retry without promoting temporary output.
How much VRAM do background removal and upscaling need?
BiRefNet uses CPU and Real-ESRGAN uses Vulkan; they are outside the generative 8 / 12 GB tier system.
Is Image Studio free?
Application code is Apache-2.0; models and runtimes retain their own upstream licenses and download requirements.
Source of facts
Answers follow the pinned source, current Release, and verification records; the review date appears in the trust line.