Skip to content
refs
REF-1937Source onlineBlueprint rev 01

Genie 3's interactive worlds, shown in action

A research launch film that proves navigable generated worlds through prompt examples, a sustained persistence test, and a fast range montage.

Maker @GoogleDeepMindFor Google DeepMindOriginal post ↗
Cut timeline · 25 shots02:23
00:0000:3001:0001:3002:00

Field notes / why it works

The indigo tile gallery builds curiosity before the product reveal at about 0:10, while the fixed prompt/result UI makes three examples easy to compare at 0:14–0:29. The largely continuous paint test at 0:45–1:11 visibly proves persistence and contains the strongest replay peak at 1:04–1:06. A burst of short world clips from 1:35 turns a single capability into a broad set of possible uses.

01

Plate 01 / The format in one breath

A calm research launch film opens on intriguing moving previews, names a new capability, then proves it through prompts, live interaction, a persistence test, a rapid variety montage, and applications. Its promise arrives in speech at 0:05–0:11: these scenes are navigable worlds generated by the product.

Why it works

  • The opening withholds a conventional product UI until 0:14: dozens of tiny moving thumbnails create curiosity while the narrator states the promise by 0:11.
  • The claim gets a visible test: at 0:58–1:10 a blue paint roller marks a wall and floor, the view moves away, and the marks remain. The 1:04.3–1:05.8 segment is the strongest replay peak.
  • Range becomes undeniable through a burst of roughly 12 scenes from 1:27–2:08, with the fastest cuts at 1:35–1:40.
02

Plate 02 / Format card

Platform · aspect · lengthYouTube · 16:9 · 2:22.6 · source 640×360 at 23.98 fps
Pace25 shots, 24 measured cuts · 10.1–11.8 cuts/min allowing uncertain changes · 17.7 picture changes/min including 18 internal changes · mean shot 5.7 s, median 3.13 s
Script291 spoken words · 124 wpm over runtime, about 191 wpm while talking · 22 sentences, mean 13.2 words · no questions
VoiceOne low-register, assured, animated narrator; direct second person
On screenAbout 70% moving product-world or POV proof, 18% dark UI/title graphics, 12% live-action operator/keyboard inserts (visual estimate from sheets)
CaptionsNo persistent subtitles; sparse centered white headlines, tiny UI labels, and prompt text inside dark cards
SoundQuiet tonal cinematic bed roughly 12.7 dB below voice; no reliably identifiable cut SFX; −14.3 LUFS, −1.9 dBFS true peak
Structure7 beats: mystery → prompt examples → live response → memory test → promptable event → variety → applications and logo; no spoken CTA

Pro / 13 plates + recipe

The first plates are open. The full dissection is Pro.

Get the timed shots, type and sound specs, and the runnable recipe.

Open the full blueprint
  1. Plate 03 / The hook, frame by frame (first 14 seconds)Pro
  2. Plate 04 / Structure (beat sheet)Pro
  3. Plate 05 / Shot grammarPro
  4. Plate 06 / Visual style guidePro
  5. Plate 07 / Voice and deliveryPro
  6. Plate 08 / Sound designPro
  7. Plate 09 / PackagingPro
  8. Plate 10 / Script templatePro
  9. Plate 11 / Production planPro
  10. Plate 12 / PromptsPro
  11. Plate 13 / Edit recipePro
  12. Plate 14 / QA checklistPro
  13. Plate 15 / Make it yoursPro
  14. Recipe / runnable promptPro