A working result. A reproducible explanation.

Prepare the computer, then the scene.

1. Use the isolated preview and a clean synthetic browser profile. Close unrelated tabs, silence notifications and keep credentials/private records off screen.
2. Set the capture to 1920×1080. Keep the browser/editor text readable and reserve the outer whitespace for later annotations. Capture individual keyframes; do not combine screens into a collage.
3. Load Studio normally. Startup may migrate valid legacy preferences or expire invalid/old state. After loading, choose Prepare temporary walkthrough, then Confirm temporary walkthrough.
4. Use Before, Operations comparison, Rejected import and Reset rehearsal. Links to scenes only navigate; they never reset or import on their own.
5. Rehearse SAY / SHOW / EXPECTED in the order below. Pause on results. Use short cuts between actions rather than racing through unreadable clicks.
6. End with Restore saved workspace. It reads the latest saved settings and preserves newer changes from other tabs. Reload also leaves temporary mode; denied storage remains page-only.

The checkpoint protects valid saved settings after initialization. It does not promise byte preservation of corrupt, expired or legacy raw envelopes before ordinary startup handling.

Nine quick cuts, result first.

Every numbered cut has an exact tutorial anchor and a scene destination. Establish the result, explain its boundary, then show code and the verification. A future video can adopt these as chapters after editing; no invented timestamps are supplied.

What to retain with the recording.

Keep the script revision, matching shot IDs, exact source revision, actual browser/version, 1920×1080 dimensions and SHA256 for every final individual frame. Include before, after, rejected import and reset/restore coverage. Keep diagnostic or intermediate captures out of the final frame list.

Use the optional presentation mode for readable capture text: this route sets code to 20px and table text to 18px at the default root size, retaining a maximum 1440px content frame for outer annotation whitespace. Check actual computed sizes and the reduced player; do not enlarge an unreadable raster afterward. Split a shot when its code/action needs a closer view.

Record the local test results separately. A frame does not establish persistence, rollback or throughput. The small CPU parse benchmark in the tutorial excludes rendering, I/O and storage; it is not a production performance result.

Future YouTube planning below validates a URL locally. It stores nothing and loads no iframe, thumbnail or tracker. Add real captions, a checked transcript and an accessible click-to-load embed only after a separately reviewed video exists.