OpenAI —
Introducing ChatGPT Images 2.0 A state-of-the-art image model that c...
- Platform
- x
- Length
- 1:00
- Views
- 12.9M
- Median cut
- 0.77s
- Face on screen
- 0%
- Dark frames
- 0%
x-bookmarks
Word-Object Montage on White
medium effortSteal this
Put the words inside the world instead of on top of it: every line of copy is a practical object in a real-looking scene, and the only graphic layer is the white canvas you reposition each image on.
Sixty seconds of roughly forty-two discrete micro-shots, each one a self-contained image in which a single word is built into the physical world of the frame (printed on a clock face, spelled in cereal, on a road sign, stamped on a toy badge). The shots are composited onto a plain white canvas at wildly varying sizes, aspect ratios and screen positions, so the frame itself becomes a layout tool rather than a fixed window. A recurring green chameleon character appears in nearly every scene as the visual thread that stops the montage feeling like a stock reel, and the whole thing resolves into a centred product lockup.
A clean, sunlit wide of a suburban house with a chameleon-shaped topiary sits letterboxed on white, held long enough to read as a normal establishing shot. It sets the naturalistic, camera-real tone and plants the character silhouette, so the first hard cut to a macro clock face at ~0.8s lands as a jolt rather than a continuation.
Relentless and even: 42 detected shot changes across 60s, median 0.77s per cut, which is close to one image per beat of a mid-tempo track. Hold the first shot noticeably longer (2s or so) and the final lockup longer again (3s+), and keep everything between at 0.6-1.0s. Vary shot length by no more than a third inside the body so the rhythm reads as machine-precise; use the mosaic frames and the empty-white two-up as your only rhythmic punctuation.
There is no overlay typography at all in the body — every word is rendered as a physical object inside the shot (engraved, printed, spelled in food, painted on signage). All the graphic design happens in the canvas: pure white background, images placed as hard-edged rectangles with no drop shadow, no stroke, no rounded corners, at freely changing scales, aspect ratios and positions. Each source image keeps its own grade — bright daylight, tungsten night, pastel nursery, high-contrast studio — and the white surround is what unifies them. End card uses a thin geometric grotesque in near-black, set flush at the left and right thirds, generous letterspacing, image centred between.
- Sunlit exterior wide, house frontage, letterboxed 16:9 inside white frameEstablishing beat and character plant; slowest shot in the piece so the eye can settle before the cutting starts
- Extreme macro, clock dial filling frame edge to edge, high-key white with saturated green handsScale whiplash after the wide; first example of a word living inside an object
- Top-down flat lay, cereal bowl centred on grey worktop, spoon on the diagonalChanges camera axis to overhead so consecutive cuts never repeat the same geometry
- Medium exterior, storefront sign board with subject perched on top, daylight, right third occupied by background signagePuts the word at true optical centre while the environment carries colour
- Night medium, brown road sign lit by a hot practical flare, black surroundThe one dark-value beat; contrast valve in an otherwise bright reel
- Multi-image mosaic: five images of different aspect ratios tiled with white guttersCompresses several beats into one frame; acts as a rhythmic accelerando
- Asymmetric two-up: large image top-left, small image mid-right, lots of empty whiteUses negative space as a beat of silence; deliberate off-balance layout
- Environmental medium, child shot from behind at an aquarium, hand reaching inHuman presence without ever showing a face, keeping attention on the object and the character
- Stylised dark portrait-mode setup, hard top light, subject in costume, near-black surroundGenre-shift shot; proves the format can absorb totally different lighting looks
- Cut-out subject on flat sky, pushed hard into the right third of a white canvasExtreme composition offset; makes the following centred shot feel like an arrival
- Macro, glitter badge on a nursery rug, shallow depth, pastel paletteTexture and tiny-type payoff shot near the end
- Centred small image with black wordmark set left and right of it on whiteProduct lockup; the layout logic of the whole video resolved into a title card
- Write your line first, then assign one word per shot; you need roughly one word per 0.8s of runtime.
- For each word, invent a physical object that could plausibly carry it in a real scene — dial, sign, badge, food, packaging — and note the lighting world that object lives in.
- Pick one recurring character or prop and place it in almost every shot; this is the only continuity device, so do not skip it.
- Generate or shoot each beat as a separate clean image or 1-2s clip. Vary axis deliberately: wide, macro, overhead, behind-shoulder, cut-out.
- Grade each beat to its own world rather than to a single house LUT — the white canvas will do the unifying.
- Build a white solid as the bottom layer of the timeline and place every asset on top as a rectangle, never full-frame.
- Change scale and position on every cut. Include at least one mosaic of 4-6 tiles and one deliberately empty asymmetric layout as rhythm breaks.
- Cut on a fixed grid at 0.77s. Do not use transitions; hard cuts only.
- Hold the opening establishing shot about 2s and let the last shot run 3s+ into the lockup.
- Finish with the product name set in a thin grotesque, flush left and right of a small centred image, on the same white.
- Generative image/video tool for producing the individual word-in-scene beats (or a props kit plus label maker if shooting practically)
- Camera capable of macro if shooting for real: 60-100mm macro plus a 24-70 for the wides
- One hard key with a bounce card — most beats are single-source, high-key daylight
- Small LED for the single night/practical-flare beat
- Timeline editor with precise transform controls (scale/position keyframes) and a white solid as base layer
- Beat-grid or metronome reference to lock the 0.77s cut cadence
- Consistent character asset or prop so continuity survives 40+ shots
- All audio: whether there is voiceover reading the words, music, or diegetic sound design cueing the cuts
- Whether the cut cadence is locked to a music beat
- Whether shots contain internal camera or subject motion, or are stills with slow pushes
- Whether assets were generated or practically shot and retouched
- Exact typeface of the end card and whether it animates in
Written by anthropic/claude-opus-5 from the frames and measurements above. Technique only, and it cannot hear the video.
Nothing matches that combination.