# Type a four-word title on, one word per beat, stepping down and around a centred phone

Manual DaVinci Resolve / Fusion construction guide generated from the recipe. It is not a tested .drp.
The accompanying Remotion project (recipe.json) is the executable definition; the numbers below are copied from it.
Neither construction establishes source fidelity: this is an original vector approximation of an inspected editing moment.

## Source and clock
Reference: OpenAI's "Image interaction with GPT-Live" (KS4X-BPjxN8), published 2026-07-08.
Source interval [0, 80) on the 24000/1001 fps master: 80 frames, 3.336667 s. Local frame 0 = source frame 0.
Create a 1920x1080 progressive timeline at 24000/1001 fps and a Fusion composition exactly 80 frames long. Keep integer frames; do not round the frame rate.
Background field: #12100e. Export silent H.264 yuv420p to match the intended playback container.

## Observation
A hand holds a phone with a blank chat screen against a dark room with warm light streaks; a fingertip hovers near the composer. Large white words type on character by character in a staircase: 'Use' at upper left (local ~7-11), 'images' below it (~16-20), 'with' to the right of the phone lower down (~25-31) and 'Voice' bottom right (~35-39). The four words hold; at ~64-72 a soft lilac orb blooms on the phone screen, and the shot cuts away at master frame 83.

## Mechanism
Word-per-beat staircase title: each word types on with a fast per-character reveal at its own fixed position, stepping diagonally around a device so the copy frames the product instead of covering it.

## Phases (local frames)
- 0: Phone, no copy
- 7: Word 1 types on
- 16: Word 2
- 25: Word 3
- 35: Word 4, title complete
- 64: Orb blooms on screen

## Layer inventory (absolute 1920x1080 top-left pixels, back to front)
- streak-a [rect] at (-200, 300) size 900x260 fill #cbb89a
- streak-b [rect] at (1250, 250) size 900x260 fill #cbb89a
- phone [group] at (790, 90) size 340x720 radius 48 fill #1b1b1b
  - screen [rect] at (804, 104) size 312x692 radius 38 fill #f4f4f5
  - notch [rect] at (910, 110) size 100x22 radius 11 fill #1b1b1b
  - status [text] at (830, 116) text "9:41" 13px weight 600 #1b1b1b
  - chat-title [text] at (790, 150) size 340x0 text "Chat" 13px weight 400 #6f6f6d
  - composer [rect] at (820, 720) size 280x44 radius 22 fill #ffffff stroke #dcdcdc 1px
  - composer-t [text] at (852, 733) text "Ask anything" 12px weight 400 #9c9c9a
  - plus [text] at (832, 730) text "+" 16px weight 400 #5c5c5a
  - mic [ellipse] at (1068, 730) size 24x24 fill #1b1b1b
  - blob [ellipse] at (900, 390) size 120x120 fill #7f7cf0 visible [64, 80)
      keyframes opacity: f64=0, f68=1 (out)
      keyframes scale: f64=0.2, f72=1 (out)
  - blob-hl [ellipse] at (920, 408) size 60x44 fill #c9d8ff visible [64, 80)
      keyframes opacity: f64=0, f70=1 (out)
- w1 [text] at (360, 380) text "Use" 84px weight 600 #ffffff reveal chars frames 7-11 visible [6, 80)
- w2 [text] at (384, 520) text "images" 84px weight 600 #ffffff reveal chars frames 16-20 visible [15, 80)
- w3 [text] at (1464, 640) text "with" 84px weight 600 #ffffff reveal chars frames 25-31 visible [24, 80)
- w4 [text] at (1440, 880) text "Voice" 84px weight 600 #ffffff reveal chars frames 35-39 visible [34, 80)

## Construction steps
1. Timeline 1920x1080 at 23.976 fps; Fusion composition 80 frames. Local 0 = master frame 0.
2. Background: near-black (#12100e) with two blurred warm streaks rotated -28 degrees; phone body 340x720 radius 48 centred at x=960 with a light screen.
3. Four Text+ nodes, 84 px semibold white, positioned at (360,380), (384,520), (1464,640), (1440,880). Each uses Write On (characters) over its own range: 7-11, 16-20, 25-31, 35-39.
4. No caret and no fade: characters appear whole. Words never move after landing.
5. Orb: 120 px ellipse on the screen, Size 0.2 to 1 with ease-out over frames 64-72 and Blend 0 to 1 over 64-68.
6. Easy mistake: giving every word the same duration; the source types short words in 4 frames and longer ones in 6.

## Content variants
Default copy:
- w1: Use
- w2: images
- w3: with
- w4: Voice
- screenTitle: Chat
- placeholder: Ask anything
Adaptation changes:
- w1: Talk
- w2: through
- w3: your
- w4: photos
- screenTitle: Live
- placeholder: Say something

## Limits
Easing names map to Fusion spline shapes: linear = linear, in/out/inOut = ease curves on the same frames, hold = step, spring/overshoot = slight overshoot past the target before settling.
Text reveal frames give the first and last unit; distribute characters/words evenly between them. Cursor click rings last 14 frames.
Vector illustration is an approximation of layout and motion, not the product UI or recovered production assets.