Kimi K3 Sketch-to-World Launch Film
A 57-second launch film turns a hand sketch into interactive scenes, then accelerates through games, 3D objects and research outputs.
Field notes / why it works
The 0:00.33 cut from a strange black-on-white pattern to an empty lightbox opens a visual question, and the 0:02–0:17 sketch, interface and rendered-world chain answers it with proof. Two specific refinement requests at 0:17–0:21 make the workflow legible; the roughly one-to-three-second examples at 0:27–0:39 show range without a feature-list voiceover. A slower research and dashboard gallery at 0:40–0:51 extends the promise beyond games before the spare brand close.
Plate 01 / The format in one breath
A product launch film begins with a strange analogue image, reveals a creation interface, follows one rough idea into a finished interactive world, then races through varied outputs before a spare brand sign-off. Its promise becomes explicit around 0:14.7 with the product name; the opening instead asks the viewer to resolve the connection between sketch, prompt and result.
Why it works
- A magnetic-looking drawing at 0:00 is replaced at 0:00.33 by a bare lightbox. The delayed reveal creates a visual question without a narrated setup.
- The sketch-to-world progression at 0:07–0:21 gives the capability a causal demonstration: drawing, interface, code, scene, then an iterative text request.
- The 0:27–0:39 run changes examples about every 1–3 s, moving from a handheld game to physics, RPG, FPS, racing, voxels and 3D objects. The breadth is concrete rather than stated as a feature list.
Plate 02 / Format card
| Platform · aspect · length | X · 1280×720, 16:9 · 56.7 s at 29.97 fps |
| Pace | 23 shots; 22 certain cuts, 23.3 cuts/min; up to 30.7/min if seven uncertain changes are cuts. Mean shot 2.47 s, median 1.42 s. Four additional internal picture changes; about 27.5 certain picture changes/min. |
| Script | No reliable narration. The measured 22-word duplicate at 0:25–0:31 has low confidence and appears to be a transcription false positive (inferred). Use short on-screen UI prompts and labels. |
| Voice | None required; no presenter or talking head. |
| On screen | Approx. 40% product/interface or graphic composition, 35% created 3D/world examples, 15% physical/analogue desk material, 10% title/end graphics (estimated from sheets). |
| Captions | No dialogue captions; small cream rectangular labels on demos and sparse black editorial serif titles. |
| Sound | Continuous music-like bed is suggested by the audio measurements (inferred); rough pulse 101 BPM, confidence 0.45. No clear evidence of cut-by-cut effects. Measured −10.8 LUFS integrated, reported +3.0 dBFS true peak. |
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- Plate 03 / The hook, frame by frame (first 3 seconds)Pro
- Plate 04 / Structure (beat sheet)Pro
- Plate 05 / Shot grammarPro
- Plate 06 / Visual style guidePro
- Plate 07 / Voice and deliveryPro
- Plate 08 / Sound designPro
- Plate 09 / PackagingPro
- Plate 10 / Script templatePro
- Plate 11 / Production planPro
- Plate 12 / PromptsPro
- Plate 13 / Edit recipePro
- Plate 14 / QA checklistPro
- Plate 15 / Make it yoursPro