HushBrief — demo production kit
Planned footage: 2:24 (144 seconds), leaving headroom below the 3-minute cap for startup and slow responses. The recorder checks actual duration and refuses to call an overlength video successful. All sample identities are fictional. Show only the app, never account consoles, debug tokens, private configuration or real client notes.
Canonical English narration and captions
demo-narration.json is the complete English narration, with timed shots and live-mode overrides. The recorder uses that exact text for the visible subtitle overlay, narration.txt and captions.srt. Because Playwright records the rendered overlay, the English captions are burned into the video pixels, without a separate caption-upload requirement. There is no synthetic voice. The same text can be read as an optional owner voice-over.
The 12-shot list follows. Timings are planned; recording-report.json records actual timings and output duration.
| Planned time | Shot | Required observable result |
|---|---|---|
| 0:00–0:12 | Product and specific audience | Clear rehearsal or configured-workspace banner; no implied live inference |
| 0:12–0:28 | Original notes beside cloud boundary | Email, phone and private passage masked; human-review limit explained |
| 0:28–0:40 | Create with cloud off | Actual result-source badge; rehearsal uses local rules, not a pretend Nano success |
| 0:40–0:52 | Action cards and evidence | Source quotes, unassigned owners/dates and review requirement |
| 0:52–1:04 | Permit reviewed preview | Explicit consent and excluded original private passages |
| 1:04–1:20 | Run cloud-capable route | Fake-cloud badge for rehearsal; genuine Gemini cloud badge required for live mode |
| 1:20–1:30 | Edit notes | Permission revoked and previous export disabled |
| 1:30–1:42 | Disconnect browser context | Actual offline local result, no system-network changes |
| 1:42–1:54 | Recovery | Labelled fake-cloud failure and attempted-route explanation; live cut continues reviewing offline recovery |
| 1:54–2:06 | Export Markdown | Real file download with processing source and review reminder |
| 2:06–2:16 | Clear workspace | Input, consent, cards and export state reset |
| 2:16–2:24 | Product closing | Accurate validation status; no unfilled repository-link promise |
Record the fake-cloud rehearsal
npm run build
npm run rehearse
In another terminal:
npm run record:demo
It creates a unique output/demo/rehearsal-* folder with hushbrief-captioned.mp4, capture.webm, captions.srt, narration.txt, cloud-frame.png, a sample Markdown export and recording-report.json. Nothing is uploaded. A fresh temporary Chrome profile is created and removed after closing. Other profiles are never inspected. Only project-local artifacts are produced.
Use npm run record:demo -- --quick for a roughly 20-second smoke check. Its narration is deliberately too compressed for a submission; use the full recording for review. The first run may download Playwright's small video encoder into .cache/playwright-video. MP4 encoding uses FFmpeg from PATH, FFMPEG_EXE or installed Python imageio-ffmpeg.
Re-run against genuine Firebase
Complete the owner command card, including a real cloud result, enforced App Check and production domain attestation. Then:
npm run record:demo -- --live --headed --url https://YOUR_DEMO_PROJECT.web.app
This selects Firebase, uses live narration and refuses a fake, local fallback or device-only result in the cloud scene. On a Nano-ready environment the SDK may choose the device; use a genuinely Nano-unavailable environment for this cloud scene. It does not manipulate capability detection in live mode. Production hosted attestation avoids registering another private localhost debug token for the new recording profile. Automation can itself fail attestation; fix/validate that environment rather than relabelling a fallback.
The current machine reports Nano unavailable. If eligible hardware later produces a genuine Nano SDK response, capture that real badge separately and replace the local-rules scene in an edited cut, keeping the final video under 3 minutes. Do not insert simulated Nano as live proof. A rehearsal-only video cannot establish the required live integration.
Optional voice-over hook, using your own audio:
npm run record:demo -- --live --headed --url https://YOUR_DEMO_PROJECT.web.app --voiceover output/voiceover.wav
Audio is padded or trimmed to the video length. No voice is created now. Inspect the whole resulting MP4, check caption legibility and actual source labels, upload the live cut to YouTube Public yourself, and verify it logged out before submission. The official rules require English or English subtitles and a public video; they do not require a face or personal voice.