Norwegian reference for local AI — registers, plans and honest tests.
Surface Book 2 as a local AI node: HyperFrames?
The problem
You have an older laptop and want to know whether it can work as a local AI node — especially for video work — without sending everything to the cloud.
The machine
Microsoft Surface Book 2, 15-inch. An older laptop that now runs Linux. The purpose here is to use it as a local AI node. RAM and GPU figures are not verified in this test; any GPU claims are treated as unverified until they are measured on the machine.
The test (tools and cost)
HyperFrames is an open video tool from HeyGen (HTML → MP4, CLI and Studio) built for AI agents: clipping, preview and rendering. Studio is said to run on Mac, Windows and Linux. In theory this node can run HyperFrames locally for video generation, clipping and video work.
Cost: not measured. The machine is not set up for testing yet — no installation has been finished here.
Planned method
When the machine is ready, we measure only what we actually get running:
- Setup: Linux environment, Node/FFmpeg (requirements according to HyperFrames), whether CLI or Studio starts without errors.
- Whether HyperFrames runs: preview and one local render of a small HTML project.
- Time per clip: wall clock for a short, fixed task (same source every time).
- Stability: does the job finish, crash, or fall back to CPU/cloud?
No tokens/s, no prices and no model names are added before they are measured on this machine.
The result
not tested yet
Results come when the test has been run. No speed figures, no cost figures and no “worked / did not work” yet.
Recipe (when the machine is ready)
- Confirm that the Linux environment has Node.js and FFmpeg available.
- Follow the official HyperFrames documentation (CLI: init → preview → render, or Studio).
- Make one small, fixed HTML project and run the same task several times.
- Note whether rendering happens locally, time per clip, error messages and whether the GPU path is active (independent verification).
- Publish numbers only after they are measured — not before.