# Camera Moves Whitebox Kit **A 132-clip whitebox reference library** for shot planning, camera-move alignment, and previsualization. > **The whitebox figure is a reference object for camera motion only.** It does not > represent the final character's look, materials, or wardrobe. Its purpose is to let > you explain "how the camera moves in this scene" *before* you shoot or build anything. **Author: hong hou zi** [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE) **Languages:** [中文](README.md) · **English** ![Camera Moves Whitebox Kit — portrait overview](docs/overview-portrait.png) > The overview is a generative illustration, not a video screenshot.

📱 Portrait one-pager (9:16, for mobile / social sharing)

![Camera Moves Whitebox Kit · 9:16 portrait](docs/overview-social-9x16.png)

--- ## Download The video files are large (111 MB), so — following GitHub convention — they live in **Releases** rather than in the repository itself. **➜ [Go to Releases to download the full pack](../../releases/latest)** | Download | Contents | Size | |---|---|---| | `camera-moves-whitebox-132.zip` | 132 MP4 clips + shot list + contact sheet + verification reports | ~115 MB | | Files in this repo | Shot list, index, contact sheet, verification reports (viewable online) | ~5 MB | --- ## Specifications | Item | Value | |---|---| | Clips | **132** | | Resolution | 1920 × 1080 | | Frame rate | 24 fps | | Duration | 4.0 s per clip (96 frames) | | Codec | H.264 / MP4 | | Scene | Checkerboard ground (2 m grid) + ring of reference pillars + near pillars + whitebox subject | | Subject | `liu_whitebox_male_default.fbx` (standing / walking / running) | > **Why are there pillars and a grid floor in the scene?** > In a plain empty scene, a dolly, a pan, and a truck shot all look nearly identical — > **without references you simply cannot tell that the camera is moving.** > The checkerboard ground and reference pillars provide parallax cues so every > camera move is clearly readable. This is what makes the library actually usable. --- ## Contents at a Glance (132 clips) | Group | Clips | Contents | |---|---|---| | 1. Fundamental moves | 8 | Dolly in / dolly out / pan / tilt / truck / crane up / orbit / follow | | 2. Classic film language | 12 | Dolly zoom (Hitchcock), zoom in/out, crash zoom, whip pan, Dutch angle, barrel roll, spiral up, high descend, low-angle push, handheld | | 3. Advanced moves (static subject) | 16 | Jib advance & rise, boom over, arc inward/outward, bullet time, FPV dive/reveal, dolly-in + tilt, dolly-in + Dutch, push-pull, impact shake, handheld lateral, rack focus, snap zoom out, vertigo + roll | | 4. With subject locomotion | 12 | Subject walks/runs via Mixamo motion (walk 4.2 m / run 9.8 m); camera tracks or stays fixed: follow behind, lead front, side track, orbit, follow rise, walk past, reveal pullback, run chase, run FPV, run side, approach, diagonal follow | | 5. Cinematic shots | 24 | Kubrick push, Spielberg push, God's eye, worm's eye, high angle, low-angle hero, overhead rotate, fisheye, telephoto, deep focus, wide low, macro close, push through frame, twist around, corridor advance, rise reveal, reverse angle, trunk shot, jump cut, face close low, pull back & up, pull out reveal, slow creep, snap push & hold | | 6. Cinematic shots, vol. 2 | 12 | Hitchcock zoom (variant), piro reveal, time slice, proscenium, flat symmetric, vertigo spin, dolly spin, over-shoulder, through pillars, tilt-up reveal, deep rack focus, scale reveal | | 7. Fast shots | 16 | High-speed motion (with motion blur; measured **3.1×–50.6× faster** than regular clips): fast truck, whip snap, snap zoom in/out, fast pass, crash zoom, double whip, rapid dolly, dive bomb, rocket rise, side wipe, hyperspeed, slingshot, barrel burst, zip lateral, strobe run | | 8. Practical shot grammar | 32 | The shot grammar you actually use on set — see below | ### Group 8: Practical Shot Grammar (101–132) This group targets the **shot grammar you actually use when shooting** — how you cover a scene, how you cut dialogue, how you transition between scenes, and how you punctuate emotional beats. | Subgroup | Shots | |---|---| | **Coverage & narrative** | Establishing pull, master wide, medium push, close-up push, insert detail, reaction drift, eyeline match, reverse pair | | **Dialogue coverage** | OTS clean, OTS reverse, dirty two-shot, conversation arc, reaction pan, group scan | | **Transitions** | Match-cut push, invisible cut, clean swish, wipe reveal, whip out, hold for dissolve | | **Subjective immersion** | POV walk, POV look around, blind reveal, door crack, negative space, diagonal frame | | **Rhythm & emotion** | Breathing push, tremor hold, heartbeat zoom, slow ramp, speed ramp, freeze push | --- ## Full Shot List Per-clip filenames with Chinese and English names: **[SHOT-LIST.md](SHOT-LIST.md)** (Chinese) · **[SHOT-LIST.en.md](SHOT-LIST.en.md)** (English). Machine-readable index (frame counts, fps, duration, file size per clip): **[manifest.json](manifest.json)** Overview image: **[contact-sheet.png](contact-sheet.png)** (132-cell contact sheet) --- ## Quality Verification All 132 clips passed the following checks; reports ship with the pack (see [`verification/`](verification/)): | Check | Method | Result | |---|---|---| | Technical specs | Independent re-verification via ffprobe | 132/132 are 1920×1080 / 24 fps / 96 frames / 4.00 s | | Framing | Camera-projection computation of subject's on-screen footprint | 132/132 subject fully in frame throughout (close-ups judged by local region) | | Motion | Dense frame sampling of the final MP4, pixel-difference comparison | 132/132 confirmed genuine camera motion | | Frame validity | Luminance / pixel std-dev / subject bright-pixel ratio | 132/132 — no black or empty frames | | Gait quality | Arm-swing amplitude / left-right anti-phase / stride | 12/12 (locomotion group) | | Foot contact & sliding | Support-foot world residual / ground gap / hip bob | Walking 9/9 pass; 3 running clips limited by the source motion file | | Speed target | Frame-to-frame pixel difference vs. slow clips | 16/16 (fast shots, 3.1×–50.6×) | --- ## Known Limitations - **Slight foot sliding on running clips**: the source motion file `Running Forward.fbx` is internally inconsistent — the ankle and toe of the *same* foot imply speeds about **0.6 m/s** apart (i.e. the foot skates during ground contact). No single travel speed can satisfy both feet, leaving a residual of roughly 0.011 m/frame. The walking clips do not have this problem. Eliminating it entirely would require a different running motion or per-foot IK correction. - **Fisheye shot (059)**: EEVEE **silently falls back to perspective** for `PANO`/`FISHEYE` projections, so that clip was rendered with Cycles. The result is correct, but re-rendering that clip takes considerably longer than the others. - The whitebox is an untextured model — **no facial features or wardrobe detail**. This is intentional: it keeps attention on the camera movement itself. --- ## How to Use 1. Download the full pack from [Releases](../../releases/latest) (or cherry-pick clips) 2. Find the shot you need in the [shot list](SHOT-LIST.en.md) 3. Drop it straight into your NLE, or use it as a blocking/previs reference Each clip is 4 seconds, self-contained, and has no audio track. --- ## License Released under the **MIT License** — free to use, modify, and distribute, including commercially; **attribution is the only requirement**. See [LICENSE](LICENSE). --- *Made by **hong hou zi**.*