- Can I get Copilot+ PC features on Linux?
- Yes, the main ones: live captions and voice typing (vinoWhisper), Windows Hello face unlock (vinoAuthFace), Studio Effects for your webcam (vinoOpenEffects), Click to Do and clipboard history (vinoPop), natural voices (vinoSpeak) and semantic search (vinoSearch). All run offline on the Intel NPU.
- Do I need a Copilot+ PC?
- No. Any Linux laptop with an Intel Core Ultra NPU works, from Series 1 (Meteor Lake) onward, including chips below the Copilot+ 40 TOPS bar. Face unlock also needs an IR camera.
- Does anything go to the cloud?
- No. Your audio, camera, face data, text and files stay on your laptop. Models download once and are checked against pinned hashes. No account, API key or telemetry.
- Is it free?
- Yes. Every project is free and open source: MIT, except vinoOpenEffects, which is GPL-3.0. vinoAuthFace's face models, from InsightFace, are licensed for non-commercial research use only.
- What is an NPU?
- A neural processing unit: a small chip beside the CPU and GPU that only runs neural networks, at a few watts. Every Intel Core Ultra laptop since 2023 has one. On Windows it powers the Copilot+ PC features; on Linux, vinoPilot uses it for the same jobs.
- Is the NPU faster than the GPU or CPU?
- For these models, yes. vinoWhisper transcribes 12 seconds of speech in 0.70 s on the NPU, 0.95 s on the Intel iGPU and 2.30 s on the CPU. It won't replace a GPU for games or very large models; it's the right size for always-on ones, and it leaves the other two chips free.
- Does running AI on the NPU save battery?
- Yes, that's what it's for. vinoSearch indexes about 5 times faster on the NPU than on the CPU at about a tenth of the energy per chunk, and indexed 9,763 files for about 10% of a charge. Idle, the NPU powers down 100 ms after its last job, so an app that is waiting costs nothing.
- Why do the apps use bigger models on the NPU?
- Because they can afford to. vinoAuthFace picks a 175 MB ResNet50 face model on NPU installs instead of the 13 MB MobileFaceNet it uses on the CPU, for a wider match margin. Studio Light's Auto light in vinoOpenEffects, a low-light model, runs only on the NPU: 10 ms a frame there, 850 ms on the CPU.
- Does Linux use the Intel NPU?
- Intel’s open source driver supports every Core Ultra NPU on Linux, but almost no desktop apps use it. The vinoPilot projects do, through OpenVINO, which keeps your CPU and GPU free.
- What about AMD, Qualcomm or NVIDIA hardware?
- Their NPUs and GPUs aren’t used yet. The apps run on OpenVINO, which only accelerates Intel hardware, so on those laptops most projects run on the CPU: slower and without the battery savings. Older Intel laptops without an NPU can use the Intel GPU in most apps. Supporting NVIDIA and AMD graphics means adding a second engine to each app (Vulkan for vinoWhisper and vinoPop, CUDA for vinoOpenEffects). Each app has an open GitHub issue for it (vinoWhisper’s (opens in a new tab), for example); a thumbs-up there decides which gets built first.
- Does it work on Bazzite, Silverblue or Bluefin?
- Yes. Every app works on Fedora Atomic (Bazzite, Silverblue, Kinoite, Aurora, Bluefin) and openSUSE Aeon. The apps install into your home directory or /usr/local, so most need nothing layered. Where a host package is needed, usually the NPU driver, the installer or doctor prints the rpm-ostree or transactional-update line to run, and ovfetch adds the NPU compiler without root.
- How do I install and update the apps?
- Each app installs with one command that downloads its latest GitHub release and checks it against the release's SHA256SUMS. After that, each one updates itself: Check for updates on its About page or in its tray menu, or its -upgrade command, such as vinowhisper-upgrade. The commands are on the apps page.