Live Stream — The 16:9 Scene Grid
One canvas, one safe area, seven named zones — every scene is built on this
162Live stream scenes
The 16:9 broadcast frame: one canvas, eight prebuilt scenes, and twelve overlays. Where the camera, the shared screen, the title, the watermark and the handles sit — and how long each one stays on screen.
The canvas contract
Non-negotiable, and set once in the encoder before the first stream — not per session. Every scene, overlay and shared source is authored against these numbers.
CANVAS | 1920 × 1080, 16:9, always | Base canvas AND output resolution. 1280×720 is the only permitted fallback, and only when upload bandwidth forces it — the ratio never changes. |
SAFE AREA | Title-safe 90%, action-safe 93% | All readable text inside 90%. The platform's own chrome, the chat panel and a phone's rounded corners all eat the outer band. |
GRID | 12 columns × 8 rows | Every zone snaps to it, in every scene. This is what makes eight different scenes feel like one broadcast. |
FRAME RATE | 30 fps for teaching, 60 only for motion | A screen share at 30 fps looks identical and leaves the bitrate for detail instead of duplicate frames. |
BITRATE | 4500–6000 kbps video, 128 kbps audio | Text-heavy screen shares need the top of that range: hairlines and 1px borders are the first thing compression eats. |
AUDIO | Mono, −16 LUFS, one mic | Mono, because half the audience has one earbud in. Audio is the only part of a stream a viewer will not forgive. |
SCENES | Eight, built before the stream | If a moment needs a ninth scene, it needs an edited video instead. |
One canvas, seven zones
The frame is built once and then held still. Of the seven zones, exactly one — the stage — is allowed to change during a session. The kicker, live badge, watermark and handle stay nailed to the same pixels from the first second to the last, which is what lets a viewer who joins at minute 40 orient in under a second.
The eight scenes
Built and named in the encoder before you go live, in this order, so the hotkeys are muscle memory. A stream that improvises a layout is a stream that spends its first ten seconds looking broken.
Live Stream — The Eight Scenes
The full scene list a teaching stream is allowed to use, and when each one is on screen
The overlay kit
Twelve graphics, and nothing else may go on a live frame. Each one has a fixed position, a fixed size and — the part that is always forgotten — a fixed duration. An overlay with no exit rule becomes furniture, and furniture is what makes a frame look busy.
Live Stream — Overlay Kit
Every graphic allowed on a live frame, with its position, size and how long it stays
Screen share, slides and the 16:9 discipline
The single most common way a teaching stream looks amateur is a source that does not match the canvas — a 4:3 org window stretched wide, or a deck exported at 4:3 and letterboxed by the platform. Neither is ever acceptable: match the ratio, or pillarbox on brand navy.
Live Stream — Screen Share & Slides at 16:9
How a shared screen, an org window or a deck reaches the canvas without being stretched or unreadable
Camera & framing
Do
- Eyes on the upper third line, a hand's width of headroom — the same crop the course thumbnails use.
- 16:9 cam, hard corners, one fixed size for the whole session (17–22% of canvas width).
- Key light slightly off-axis, background two stops darker so the cam separates from the stage behind it.
- Look at the lens for the open, the promise and the close. Look at the screen while demonstrating — pretending otherwise reads as shifty.
Never
- A circular or pill-shaped cam. Sharp geometry is Principle 4, and a round cam is the tell of a stock template.
- Moving the cam between scenes to “make room” — re-park the content instead.
- A blurred or virtual background: it eats fingers when you gesture, and gestures are half of teaching.
- Cam over the shared screen's left navigation or over the record you are editing.
The mark in the frame
- Bottom-right, 55–65% opacity, always on. The one graphic with no exit rule — a clip of your stream travels without a title, and the watermark is what still names the teacher.
- Mark + wordmark on wide scenes; mark alone when the camera occupies that corner.
- Height about 3.5% of canvas — roughly 38px at 1080p. Large enough to survive a vertical crop for Shorts.
- It animates once, in the intro sting, and never again. A pulsing logo is a distraction the viewer cannot dismiss.
- It never sits on a busy region of the shared screen — that is why the bottom-right 22% is kept clear.
- It is never recolored per session, per topic, or for a festival. See Brand → The mark.
Run of show
T−10 min | Starting soon. Countdown, session title, handles. Music optional and quiet; no talking over it. | scene 01 |
0:00 | Open on camera. The payoff first — what will exist by the end of the hour. No “can everyone hear me.” | scene 02 + lower third, 8s |
0:02 | The promise and the map. The 3–5 steps, stated aloud and put on the topic bar as step 1 of 5. | scene 04 + topic bar |
0:05 | Teaching blocks. One step per block, each closed by doing the thing. Update the topic bar at every boundary. | scenes 03 / 04 / 05 |
every ~15 min | Re-orient. Name where you are for the people who just joined, and take one question. | scene 07 + question card |
last 10 min | Q&A proper. One question on screen at a time, attributed. Unanswered ones become next week's cold open. | scene 07 |
last 2 min | Close. Recap the promise in three lines, name and date the next session, one CTA. | scene 08 |
after | Hold the end card 60s before you stop the stream — the platform's replay opens on it and clip-grabbers land on it. | scene 08 |
Social — before, during, after
BEFORE | One announcement asset, 24h out, reusing the course thumbnail frame with a // LIVE · THU 19:00 IST kicker. | thumbnails page · story template |
DURING | The handle bug rotates one destination at a time — site, YouTube, LinkedIn — roughly every 10 minutes. Never a row of five icons. | overlay 06 |
DURING | The CTA sting fires at most twice an hour, after a completed demo, never mid-explanation. | overlay 08 |
END | The ending scene is geometrically the same asset as the YouTube end screen, so the VOD and the stream close identically. | promo page · end screen |
AFTER | Two or three vertical clips cut from the recording — the watermark and the 90% safe area are what make a 9:16 crop survive. | video page · instagram page |
AFTER | The session becomes the week's Monday lesson: same title, same thumbnail family, tighter edit. | video series & schedule |
Accessibility, live
- Platform auto-captions on, always. They are imperfect and still better than nothing for the third of viewers watching muted.
- Say what you are doing as you do it — “I'm clicking Setup, then Object Manager.” The narration is the audio description.
- Never point with color alone: “the red one” excludes; “the Delete button, top right” does not.
- Question cards and the topic bar carry the same contrast floor as the product — 4.5:1 for body-size text on the navy canvas.
- No flashing transitions. The scene wipe is 320ms linear, once, and that is the whole motion vocabulary.
- Read every on-screen question aloud before answering it — chat is not visible in the VOD.
The failure modes
What makes a stream look improvised
- A scene built live, on air, because the moment needed one.
- The camera in a different corner in every scene.
- A stretched or letterboxed shared source.
- Three overlays on screen at once, none of them expiring.
What loses the viewer outright
- Type below 16px on a shared screen — most of the audience is on a phone.
- Silence during a long operation. Narrate the wait or cut to camera.
- Chat scrolling live in a panel, competing with the thing you are teaching.
- Stopping the stream the instant you stop talking, with no end card.