TASK GUIDE + WORKING TOOL / UPDATED AUGUST 2026
Video to prompt, done shot by shot
Most video-to-prompt tools hand back one paragraph for the whole clip. That paragraph cannot rebuild a multi-shot edit. This page runs the other workflow: detect the cuts of a short video you own or are licensed to analyze, locally in your browser, then write one editable, model-neutral prompt per shot for Veo, Kling, Runway, and other video models.
To turn a video into prompts, run cut detection on a clip of up to 60 seconds that you own or are licensed to analyze, then write one model-neutral prompt per shot covering subject, action, framing, camera movement, composition, lighting, color, and pace. The workspace below detects cuts locally in your browser, drafts prompts only with a separate cloud AI consent, and copies or exports the finished map without leaking file details.
What leaves this page, and what never does
Three boundaries hold the whole workflow together. You start with a video you own or are licensed to analyze. Cut detection runs locally in your browser, so the file is not uploaded to build the shot map. Cloud AI is a separate step that asks for its own explicit consent after the local result exists, and it can only fill fields, never move your cuts.
Rights and capability boundary
Prompts describe only what the shots show. Nothing here names a real person or brand, imitates a living artist, or invents events outside the frame, and copied or exported data carries no filename, account, or task details.
A Prompt Map is per shot, not one paragraph
Each row of a Prompt Map keeps the evidence next to the prompt: the shot number, exact in and out timecodes, duration, shot size, camera movement, and story beat, plus the opening frame as a thumbnail or color block. The prompt itself stays editable, so an AI draft is a starting point you correct, not a locked answer.
The example below is a hand-written, illustrative example: three shots from our labeled vertical-drama cold open. It is demo data with rights-safe wording, not model output and not footage from any real work.
ILLUSTRATIVE EXAMPLE / NOT MODEL OUTPUT
Three shots, three prompts.
Color blocks stand in for opening frames in this example. The real workspace below shows detected thumbnails.
Hand-written illustrative data, rights-safe by design, also used on the product demo path. Not model output.
Five steps from video to prompt map
Step 1: Confirm your rights to the reference clip. Choose a video of up to 60 seconds that you own or are licensed to analyze. StoryTeardown asks for this confirmation before anything runs, and it is not a downloader or library for other people's videos.
Step 2: Detect the cuts locally. Cut detection runs locally in your browser. The file is not uploaded for this step; the browser samples frames and marks frame-change peaks as candidate cuts.
Step 3: Review the shot timeline. Check the detected boundaries against the picture, then correct shot size, camera movement, characters, action, and story beat per shot. Every field stays editable.
Step 4: Optionally add one cloud AI pass. With a separate, explicit consent, one cloud AI pass drafts the craft fields and a model-neutral prompt for every shot. It cannot change your cut timeline.
Step 5: Edit, copy, and export the prompt map. Rewrite any prompt, copy a single shot or the whole pack, or download the CSV. Copied and exported data carries prompts, timecodes, and craft fields only.
What copy and CSV carry
Copy one shot, copy the whole pack, or download the CSV: the payload holds prompts, timecodes, and craft fields only. Filenames, accounts, task ids, and other source-video details never enter the export, and CSV cells are prefixed so spreadsheet apps cannot mistake prompt text for a formula.
Review the privacy policy and terms before using cloud analysis.
FAQ
Can I turn any video into prompts?
No. The workflow is built for video you own or have explicit permission to analyze, and it asks you to confirm that before analysis starts. It does not fetch videos from platforms or host a public library.
Does this upload my video?
No. Cut detection runs locally in your browser and sends nothing to a server. Cloud AI drafting exists as an optional step behind a separate, explicit consent that you give after the local result is ready.
Is this one prompt for the whole video?
No. A single paragraph cannot carry a multi-shot edit. StoryTeardown builds one editable, model-neutral prompt per detected shot, each tied to an exact timecode, the craft fields, and the shot's opening frame.
Do the prompts work with Veo, Kling, or Runway?
The prompts are written in model-neutral language: subject, action, framing, camera movement, composition, lighting, color, and pace. That wording pastes cleanly into Veo, Kling, Runway, or any other video model, and you can adapt each prompt before copying.
What do copy and CSV include?
Prompts, timecodes, and craft fields only. Filenames, accounts, task ids, and other source-video details never enter the copied text or the CSV, and CSV cells are guarded against spreadsheet formula injection.
Where to go next
PROMPT MAP STUDIO / LOCAL MVP
Start with a reference video you have the right to analyze.
Choose a clip of up to 60 seconds that you own or are licensed to use. Cut detection runs in this browser; optional cloud AI drafting is a separate opt-in, and every prompt stays editable.
THE WORKFLOW
From authorized reference video to an editable Prompt Map.
- 01Choose & confirm
Select a local short clip and confirm that you own it or have permission to analyze it.
- 02Build the shot map
Detect likely hard cuts on this device and review the fixed timeline.
- 03Optional AI draft
With separate consent, one cloud pass drafts craft fields and per-shot prompts. It cannot change the cuts.
- 04Edit & pack out
Correct any prompt per shot, copy single prompts or the whole pack, and export the CSV.
Choose cut sensitivity
Hard-cut detection only. This setting does not claim to understand the story.
RECENT WORK / THIS BROWSER
Continue a saved breakdown
Completed results stay on this device so closing a tab does not erase the delivery.
ACCOUNT, CREDITS & TASKS
Local safe skeleton — external capabilities pending verification
This external capability is temporarily unavailable or incompletely configured, so the server has failed closed. Local cut detection and exports remain available; no account, balance, payment, or cloud result is fabricated.