Verification ledger
Not meeting-tested. Native scene and local virtual-camera/BlackHole fixture checks passed on the confirmed Mac; Zoom end-to-end delivery remains untested. Date: September 5, 2026 UTC.
| Layer | Status | What it proves / does not prove |
|---|---|---|
| Official control documentation | Reviewed; links in the guide | Described controls, not the presenter's installed UI |
| Presenter machine and application versions | Confirmed / inventoried | User-confirmed Mac; macOS 26.6.2, OBS 32.0.1, Zoom 7.0.6, Chrome 152.0.7977.77, BlackHole 0.6.1. Installed bundle/API inventory, not all About windows |
| Original fixture generation and static checks | See downloadable verification.json |
File properties, checksums and separate silent presentation; not Zoom reception |
| Actual recording / timed transcript | Private material excluded | Public timings remain blank; invented example clearly labeled |
| OBS-exported collection and profile | Native configuration created; menu exports pending | Separate rehearsal collection/profile exists on Mac. No fabricated export or untested import download supplied |
| Fresh OBS import / alternate-media adaptation | Not run | Native menu interaction remains unavailable; file review alone cannot pass |
| Annotated installed-application screenshots | Not captured | Routing diagram is a schematic, not a screenshot |
| Native activation, active-scene reselection, SAFE/reset, EOF | Passed fixture checks | Three API-triggered starts; active PLAY selection did not restart; SAFE deactivated; full 30-second fixture reached finished card without looping. Not manual timing or attendee first frame |
| Native virtual-camera receiver | Passed local fixture check | Temporary native video input, never placed in Program, received READY, playback, FINISHED and SAFE. This was reception through the virtual-camera device, not a Program preview; not a Zoom endpoint |
| Native BlackHole return capture | Passed local fixture check | HEAD stayed Monitor Only; temporary BlackHole input monitoring was OFF. A 35.03-second local recording contained three diagnostic bursts at the expected 14-second and 10-second spacings, with no other audio above the stated threshold |
| FFmpeg/AVFoundation consumer harness | Inconclusive / unsuccessful | Initial camera attempts stalled/rejected frame rate and audio captures were short/silent. Native receiving-input tests above subsequently demonstrated the local routes; the harness failure is retained |
| Device-loss and meeting recovery procedures | Not run | No deliberate hardware disconnection, OS routing change or live meeting test |
| Independent-attendee Zoom rehearsal, three starts | Not run | Requires authorized meeting and endpoint |
| Full-duration pair, lip sync and shutdown margin | Not run | Requires actual footage, approved presentation and endpoint observations |
Local site checks exercised navigation from existing sections; blank, illustrative, invalid and failed timing windows; JSON download; light/dark appearance; layouts at 390, 768 and 1440 CSS pixels; two-page runbook printing; and playback of both separate fixture videos. These are website/artifact checks, not installed OBS or Zoom screenshots. Browser evidence stays local. Official link checks reached the expected OBS, Zoom and Google article titles; the BlackHole landing page returned HTTP 406 to the direct automated check but was readable through a separate web retrieval. Its official repository and install/uninstall pages were reachable.
An independent fresh-context editor unpacked the original archive, verified its checksums and full media decoding, regenerated seven picture/video assets byte-identically in an isolated environment, and exercised twelve offline worksheet cases. That review found an ambiguous first-motion timing instruction, broken Markdown-entry links and missing alternate-media preparation commands. The guide now defines media-time-zero measurement (or subtracting a later event's known source timestamp), ships offline-resolving Markdown links and gives explicit probe/one-track preparation commands. This review did not reproduce OBS imports, alternate-media application playback or a Zoom meeting. Those gates remain not run; the record of assistance and initial failures stays local.
Capture packet still required
Capture locally on the isolated, neutral-fixture setup. Annotate the selected control and its intended consequence; crop or redact accounts, meeting details, device identifiers and identifying paths. Do not substitute vendor screenshots or mock applications. Public release of private rehearsal evidence requires explicit approval.
- READY / PLAY / SAFE scene lists with unique media-source placement and Cut.
- Media properties showing loop, restart, close-file and EOF flags.
- Transform panel, canvas/output/frame rate and audio sample rate.
- Disabled global audio, HEAD monitoring and exact monitoring device selection.
- Virtual-camera Program selection and Studio Mode state.
- Zoom camera, virtual mic, Mac speaker output, selected processing and in-meeting state.
- Slides manual playback/full trim and Zoom share surface/options.
- Attendee-observed first frame, one audio path, beginning/middle/end checks, EOF and manual mute/video-off completion, with consent to capture.
Acceptance record
The presenter clarified that the work laptop uses normal macOS audio output, not headphones. A read-only check confirmed MacBook Pro Speakers as the current default output. This resolves output identification, not Zoom selection or acoustic verification: speaker-based return audio and any live-mic handoff still need an attendee-side rehearsal. No OS default or Zoom setting was changed to make this documentation correction.
Fill environment.json, timing.json, trials.csv, checkpoints.csv and
acceptance.json privately. Keep test failures and assistance visible. Each
pass needs a date, application versions, fixture/media version, observer and
specific result. Agree lip-sync, alignment and readability thresholds before
testing; do not retroactively move thresholds to fit a failed run.
The native return recording was 1080p30 with one 48 kHz stereo audio stream. At a −50 dBFS detection threshold, the three returned bursts followed the corresponding local Program flashes by approximately 72.7 ms each. This is a measurement of that local recording path, not Zoom latency, speaking lip sync, a manual reaction delay R, or an instruction to apply a −73 ms offset. The virtual-camera receiver was checked separately. Observe the actual camera, microphone and screen-share paths together at the attendee endpoint before choosing a sync correction or an acceptance threshold.
Native checks used OBS's authenticated API in the isolated rehearsal collection, with all global capture sources disabled. No driver installation or OS default audio-device change was needed. Source-render screenshots showed the neutral cards and fixture; the EOF render matched the finished-card render. These are not screenshots of settings panels. The first virtual-camera assertion ran too early; polling fixed that check. The unsuccessful FFmpeg consumer was not used as evidence of success; native receiving inputs provided the local route checks. All temporary receiving inputs were removed, READY restored, and virtual-camera and recording outputs confirmed stopped. No physical capture was introduced. Remote UI automation was denied macOS Accessibility permission; this was not bypassed. Menu export/import, Zoom settings and actual attendee observations remain open gates. Private logs, recordings and device identifiers stay local.
All pending application and attendee rows must pass before describing a setup as tested. A successful website deployment, a syntactically valid export, OBS meters, a local player, or a fresh-context documentation review is insufficient.
Artifact check results / JSON · Incomplete meeting acceptance template · Archive checksum · Dependency/license manifest