Director Chair
All groups

Camera

Shot sizes 13Angles 13Camera moves 42Rigs 6Lenses 9Focus and depth 4

Frame

Composition 13Aspect ratios 6

Light and look

Light 11Time of day 4Colour looks 13Film stock and texture 9Art styles 6

People and places

Characters 5Expressions 31Settings 3

Motion

Subject motion and speed 8

Edit

Transitions 21Edit effects 41Sound 14

Story

Hooks 46Script arcs 10Video formats 11Calls to action 5Test and publish 16

Recipes

Presets 28Prompt guides 11

Tools

BoardModelsGuidesAboutAgent skill

One canvas holds the whole process as linked cards, a virtual camera walks it step by step, a presenter cut-out talks from the bottom-left corner, two-word captions beside them.

Presets

Canvas tour with a corner presenter

The viewer has the map and the part being explained on screen at once.

Use when
Process explainers and step lists where every step has a real artefact to show.

Prompt instructions

You are my video producer. Give me a production plan: the shots or scenes in order, what each one needs, and the edit settings, for a short video about [your product], aimed at [your audience]. Technique: Canvas tour with a corner presenter. The viewer has the map and the part being explained on screen at once. How it works: One canvas holds the whole process as linked cards, a virtual camera walks it step by step, a presenter cut-out talks from the bottom-left corner, two-word captions beside them. Right for: Process explainers and step lists where every step has a real artefact to show. What we learned from tested runs: - Each step lands on one real artefact: the tool's own panel, the prompt, the output. - Measured: 2.86 moves per 10 seconds with a median hold of 2.2 seconds, calmer than a stacked board at 4.7, because each card needs a second to read. - The source's true peak is -0.4 dBTP, which our QA fails. Be concrete: numbers, seconds and settings, no theory. Ask me one question first if something you need is missing.

Paste it into Claude, ChatGPT or Gemini.

Watch out

Notes

Measured on the reference

WhatValue
Length209.8 s
Aspect9:16
Visual changes per 10 s2.86
Median segment2.2 s
First visual change7.2 s
First spoken word0.0 s
Spoken words per second3.18
Loudness, integrated-14.1 LUFS
True peak-0.4 dBTP (too hot)
Sound floor-85.7 dBFS (clean silences, no music)
Face in sampled framesnot usable: the detector counts the people pictured on the canvas
Plays, saves31,200 plays, 6.2 percent saved (read 23 Oct 2026)
Against the account's median5.9 times

The recipe

  1. Build one large canvas first: a title, then every step as a card in reading order, linked by dashed lines, each card holding the real artefact of that step.
  2. Place the presenter as a cut-out in the bottom-left corner, about 30 percent of the width and the lowest quarter of the height, for the whole video.
  3. Frame 0: the title and the full step map, readable. The first word is on frame 0.
  4. Move the camera to each card when its step is named and hold until the next: 2.5 to 3.5 moves per 10 seconds, each hold 1.8 to 2.6 seconds.
  5. Pin a dark pill with the step number and name at the top left, and show the step title large with its number in an accent circle.
  6. Zoom out to the whole canvas once near the end, then close on the map of frame 0 so the loop restarts cleanly.
  7. Captions of two words (one to three), white bold sans on a dark rounded box to the right of the presenter, the spoken word in the accent colour.
  8. No music, no sound effects. Voice at -14 LUFS.

Not for

Emotional stories, testimonials, product beauty, live-action brands, anything under 30 seconds.

Related

One large board holds every step; a virtual camera pans and zooms to the step being named and holds there until the next one.

Camera move on one canvasPrompt

The presenter as a cut-out in the bottom-left corner, about 30 percent of the width and the lowest quarter of the height, the caption chip to their right.

Presenter in the cornerPrompt

The first frame shows the entire board zoomed out, the title and all ten numbered steps, before the camera goes to step one.

The whole list as a map in frame 0Prompt

The last frame returns to the opening frame: the same map of all seven steps, the presenter in the same corner.

Loop back to frame onePrompt

Kinetic type, interface renders, product shots and transitions cut to music, with no live footage.

Type and shapes carry itPrompt

Sources

Our own reading and measurement of a TikTok by @pro.glitch, posted 22 Oct 2026. The clip is a third-party reference and is not shown; the words here are ours Source Checked 2026-11-13.

For agents: this card as markdown, as JSON.