🎬 Synexar Demo Video — Production Plan

fully autonomous AI production · July 5 · wet ink — tap any box to flag it, notes in the tray

✅ SHIPPED — v1 delivered (1:52, 1080p). Real staging footage (synthetic patient "Jordan Rivera" only — PHI-check passed on every list view), OpenAI narrator, music bed, animated brand titles, SRT captions. Sent to you as synexar-demo-v1.mp4 + production notes. Mark up this board for the v2 cut — or record your own voice over the same timeline (8 clips, notes explain the swap).

The pipeline

0 · ToolchainFFmpeg 8.1.2 + neural TTS + Playwright 1.61 — all verified on this machinedone ✓

→

1 · ResearchToolchain verified hands-on: OpenAI steerable TTS, FFmpeg 8.1, frame-rendered titles (web-research agent flaked; decisions made from verified local checks)done ✓

→

2 · Script8 scenes, ~250 words of narration → 1:52 cutdone ✓

→

3 · Capture3 shoots on staging (cases 15742–15744): tracker, consent signing, OR Studio (DICOM), live AI generation, charge ticket, GI QuIC, sign & printdone ✓

→

4 · Voice + MusicOpenAI "onyx" narrator (edge-tts fallback kept) + "Deliberate Thought" (CC-BY, credited on end card), side-chain duckeddone ✓

→

5 · AssembleFFmpeg: crossfades, push-ins, VO-timed cuts, −16 LUFS master + SRTdone ✓

The film — 8 scenes

1 · 0:00Hook"15–20 min of documentation after every scope."title card

2 · 0:16RevealSynexar logo → live Patient Trackerlogo + app

3 · 0:30Front officeIntake, consent, history — one guided flowlive app

4 · 0:48ProcedureStructured findings in a handful of clickslive app

5 · 1:06 ★Doctor StudioAI writes the full report in seconds — "4 clicks" counter overlaymoney shot

6 · 1:34BillingDeterministic CPT + NCCI — "rules, not guesses"live app

7 · 1:52GIQuICRegistry fields auto-derived — no duplicate entrylive app

8 · 2:08Close / CTA"Documentation at the speed of medicine" · synexar.netlogo card

⛔ Hard guardrails

Zero PHI in any frame — the capture script creates its own synthetic patient; no real names / MRN / DOB ever on screen.
Filmed on staging only — prod untouched.
No product code changes — this is pure filming.
AI voiceover is a placeholder track — swappable for your voice without re-editing.

✅ Where it stands right now

FFmpeg 8.1.2, neural TTS, Playwright — installed & verified
Script v1 written (8 scenes, ~250 words VO)
Tool-landscape research + staging-automation recon running in parallel
Next: film the app → generate voice → assemble

v2 options — flag what you want

Your voice Record the 8 scene lines (script in notes) → I re-master the same timeline in minutes. AI "onyx" narrator stays until then.

Cinematic b-roll intro AI-generated hospital/endo-suite shots (Veo/Sora) to replace the kinetic-type hook — needs a video-gen API account. Say go and I'll wire it.

60s + 30s cuts LinkedIn / teaser edits from the same footage — cheap now that the kit exists.

🐛 Real bug found while filming /case/new: picking a physician NPI suggestion pollutes the patient-name field. Issue being filed — fix routes through /synexar-task.