How to record real lab demonstrations

Português · Back to workshops

This plan completes media without fabricating an outcome, exposing credentials, or recording personal data. The current Lab 01 GIF is a real, compressed demonstration. Lab 02 media remains explicitly pending until a controlled session can be recorded.

Before recording

Use a separate OpenAI project with a small budget and alerts.
Close DevTools, password managers, notifications, and personal tabs.
Confirm .env.local is off-screen and ignored: git check-ignore -v .env.local.
Use short fictional text and speech with no names, email, company, phone, or location.
Record the application window only; never a terminal containing variables.
Use headphones for Lab 02 to prevent feedback.

Lab 01 — 15-to-25-second sequence

Show: short text → voice/format choice → generate → processing state → player → playback → download. Keep GIF audio muted; the text description explains the outcome. End before any dashboard or log becomes visible.

Lab 02 — 25-to-40-second sequence

Show: notice and consent → start → microphone permission → connected state → one short utterance → one response → interruption → mute → one text message → End → microphone indicator off. Never show Network, client secrets, API keys, personal transcripts, or Usage.

Export and accessibility

Prefer WebM or MP4 without audio autoplay; use a silent GIF fallback.
Initial target: maximum 960 px width, 10–15 fps for GIF, under 2 MB.
Remove pauses and redundant frames before lowering quality.
Add loading="lazy", dimensions, alt text, and a nearby text description.
Never autoplay sound. Respect prefers-reduced-motion with an equivalent static description.

Review before committing

Inspect frame by frame: no secret, token, request header, personal data, or notification may appear. Then run npm run docs:links and view the page at 360 px and 1440 px. If a real safe recording is unavailable, keep the honest placeholder.