/director

Demo Recording Checklist

Quality gates for each phase of demo video production.

Purpose: Read this when Director must validate demo readiness, playback quality, security hygiene, accessibility, perceptual quality, AI-citation packaging, or final delivery quality.

Contents:

• Pre-Recording Checklist: scenario, hook, environment, aspect, and accessibility setup
• Post-Recording Checklist: playback, content, security, captions, transcript, and output validation
• Pre-Delivery Checklist: packaging, AI-citation, multi-aspect, and reproducibility checks
• Quick Check (3 minutes): rapid go/no-go review
• Quality Scorecard (/97): scoring and release thresholds (2026 expanded)

Pre-Recording Checklist

Scenario

• Audience identified
• Purpose defined (1-2 sentences)
• Archetype selected (30s social / 60s producthunt / 90s linkedin / 180s walkthrough / 3×45s series)
• 3-second layered hook designed (visual + text + optional audio)
• One demo, one feature
• Story arc defined (Hook → Pain → Solution → Result → CTA)
• Emphasis points identified
• Duration plan ≤ 120s OR explicitly chaptered

Environment

• Using playwright.config.demo.ts
• Target aspect chosen (16:9 / 9:16 / 4:5 / 1:1) and viewport matches
• page.screencast planned with explicit size
• recordVideo retain-on-failure safety net configured
• slowMo appropriate for audience
• Resolution matches target device / aspect
• Test data seeded
• Auth state prepared (storageState)
• browser.bind() configured if MCP/CLI clients will attach

Accessibility

• Key screens validated with ARIA snapshots (if applicable)
• Overlay text meets contrast requirements (≥4.5:1 AA, prefer 7:1 AAA)
• Caption plan defined: open (burned-in) for muted-autoplay + closed (.vtt) for accessibility
• WCAG 1.2.2 (captions) plan committed for external distribution
• WCAG 1.2.5 (Audio Description) status decided — needed when visual content isn't fully narrated

Post-Recording Checklist

Playback

• Video plays to completion
• No choppy frames or flickering
• Text is readable at target resolution and at mobile size
• Hook locks attention in first 3 seconds (verify by watching with sound off)

Content

• All operations visible at demo speed
• Pacing feels natural (not rushed, not dragging)
• Overlays readable with sufficient display time
• Final result displayed long enough (1.5–2s)
• No 16:9 master reused as-is on 9:16 / 4:5 channels — re-shot or re-framed per aspect

Security

• No passwords, tokens, or production data visible
• No console errors or network errors visible
• No internal staging URLs, IPs, or feature flags exposed
• Demo data used (realistic but fictional)

Captions & Transcript

• Open-caption (burned-in) variant generated for muted-autoplay channels
• Closed-caption .vtt track generated and synced
• Transcript (.vtt + plaintext) generated
• Non-dialog cues annotated ([click], [music swells])
• Caption reading speed ≤ 17 CPS, ≤ 42 chars/line, ≤ 2 lines
• Product names and homophones manually QC'd (auto-caption WER baseline only)

Perceptual Quality

• VMAF ≥ 90 at 1080p (ffmpeg-quality-metrics --vmaf)
• PSNR ≥ 40 dB
• SSIM ≥ 0.95
• LUFS normalized to -14 (YouTube/LinkedIn) or -16 (Web), TP ≤ -1 dBTP

Output

• Naming convention followed: [feature]_[action]_[aspect]_[date].webm
• MP4 (H.264) conversion generated for universal playback
• AV1 archival master generated (if hero/archival use case)
• File size reasonable (~5–8 MB / 30s at 1080p VP9)
• Saved in demos/output/ directory with aspect subfolder

Pre-Delivery Checklist

• Demo meets original objectives
• UI matches current product state
• Scenario document saved for reproducibility
• Metadata recorded (date, version, environment, aspect)
• Output formats match distribution channels (WebM/MP4/AV1/GIF)
• Transcript .vtt + plaintext attached for external distribution
• VideoObject JSON-LD schema generated with hasPart chapters, transcript, thumbnailUrl
• Thumbnail set generated per platform (YouTube 1280×720, LinkedIn 1200×627, X 1600×900, Product Hunt 1200×1200)
• A/B thumbnail variants prepared (face-first vs product-first; default product-first for B2B / dev-tools)
• WCAG verdict recorded: 1.2.2 (captions), 1.2.4 (live, if applicable), 1.2.5 (AD)
• If multi-aspect: 16:9 + 9:16 + 4:5 variants exported and verified

Quick Check (3 minutes)

[] Hook lands in 0–3s (sound off)
[] Plays to the end
[] Followable pace
[] No confidential information
[] Clear result
[] Captions sync (spot-check 3 points)
[] VMAF / PSNR / SSIM verdict logged
[] Transcript + VideoObject JSON-LD attached
[] Correct filename, aspect, and format

Quality Scorecard (/97)

Rate each item. Most items 1–5; some items have higher max points reflecting 2026 priorities (hook, aspect, multilingual, perceptual quality).

Score Assessment

• 80–97: Ship clean — production-ready
• 70–79: Ship with minor fixes — log follow-ups
• 50–69: Ship with major fixes — re-shoot weakest scenes
• < 50: Reshoot — review from scenario and hook