Teardown — Africa's Deadliest Large Animal Is a Herbivore: 20 cuts, film + clips + every prompt
20 cuts across 2 motion tiers, 90% on tier 1. A six-minute episode, published, with every generated clip (20), every source still (22), and the exact still and motion prompt behind each cut.
One published episode opened up: the finished film, 20 generated clips, 22 source stills, and the prompt behind each of its 20 cuts.
What is in it
- film/ — the finished episode — the published cut, exactly as it went out
- clips/ — 20 generated clips — every shot before the edit, including the plain ones
- stills/ — 22 source frames — the image each clip was animated from
- PROMPTS.md — one block per cut across all 20: framing, tier, still prompt, motion prompt
- production.json — the same record machine-readable, one object per cut
Claims, and how to check them
- every one of the 20 cuts has its prompt included
check: count the blocks in PROMPTS.md, or the objects in production.json - the archive holds 20 clips and 22 stills
check: list the zip; these counts were read from it - the episode is published and watchable before you buy
check: the listing links the public cut
What it does not do
- narration and on-screen text are Korean; the prompts and all documentation are English
- clips are re-encoded for download, not master bitrate
- no guarantee of a result — model behaviour changes between versions
- not a template: this is one episode, not a reusable standard
You need already
- nothing to read it — markdown, JSON, MP4 and JPEG
- an i2v generator and your own credits, to re-run anything
clips: 20cuts: 20format: midformnarration: Koreanstills: 22tier_split: dynamic:2 / low_motion:18
Every AI video tutorial shows you a prompt that worked. None of them show you the other nineteen.
This is one published documentary episode, opened up. The finished six-minute film, every generated clip that went into it, the stills those clips were animated from, and — for each of the twenty cuts — the exact still prompt and motion prompt that produced it, next to the engine tier it was ordered on.
Not a curated highlight reel. The whole episode, including the cuts that are deliberately plain.
WHAT IS IN THE PACKAGE
• The finished film — 6 minutes, 1080p, as published
• 20 generated clips — 1080p, the raw shots before the edit
• 22 source stills — the frames each clip was animated from
• PROMPTS.md — one block per cut: framing, tier, still prompt, motion prompt
• production.json — the same record, machine-readable
WHY A WHOLE EPISODE INSTEAD OF A PROMPT LIST
A single good prompt teaches you almost nothing, because you cannot see what it was competing with. Twenty of them in sequence show you the thing that actually matters: how the same subject is held identical across cuts, how framing is specified so the engine does not reinterpret it, and how many shots in a working episode are ordered on the cheap tier.
In this episode: 18 cuts on the low-motion engine, 2 on the higher tier. That ratio is not an accident and it is most of the budget.
THREE THINGS YOU CAN READ STRAIGHT OFF THE PROMPTS
• Identity is locked with a positive count, never a negation. Prompts here say what the animal is — species, ear shape, coat, toe count. They do not list what it must not be, because a negation tends to summon exactly the thing it forbids.
• Framing is stated as what fills the frame, not as a shot-size name. Shot-size vocabulary is interpreted loosely by every engine we have measured.
• The motion prompt is split into subject, action and environment as separate lines. Collapsing them into one sentence is where drift starts.
HONEST NOTICE
• The narration and on-screen text are Korean — this episode was published on a Korean channel. Every prompt, and all the documentation, is in English. The prompts are the transferable part.
• Clips are delivered re-encoded for download rather than at master bitrate. They are reference and working copies, not broadcast masters.
• You may use these files in your own commercial and personal work, including client work, and modify them freely. You may not resell or redistribute the package itself, in whole or in part, as a product — including uploading the clips to a stock library.
• This is a record of what one episode took. It is not a guarantee of a result, and generation models change behaviour between versions — sometimes reversing it. Treat the routing as a procedure you can re-measure, not a fixed ranking.
WHERE IT SITS
This is the evidence under the packs. If you want the reusable standard rather than one worked example, AI Video Prompt Craft is the prompt grammar and routing rules generalised, and the Animal Bond Shorts Pack is the full ten-stage pipeline for this genre.