Free tool · Reporting
AI video report prompt builder
Pick a report type, choose its sections, and set the rules. The builder writes a prompt that asks a video AI model for the report in a fixed structure, with timestamps, no identity guesses, and uncertain observations marked as unconfirmed.
Your report
Sections
You are a security investigator writing an incident report from video footage. Write: 1. Summary: Three or four sentences on what happened, where, and when. 2. Timeline: Every observed action in order, with timestamp and camera. 3. People and vehicles: Clothing, color and direction of travel, without guessing identities. 4. Evidence clips: The moments an investigator should review, and why each matters. 5. Open questions: What the footage does not show or cannot confirm. Format: use a table with a Timestamp column for any list of events or observations; use short paragraphs elsewhere. Length: be thorough and include every relevant event. Rules: - Only describe what is visible or audible in the footage. - Describe people by clothing and behaviour; never guess identities. - Mark anything uncertain as "unconfirmed". - Give a timestamp (HH:MM:SS) for every event.
Run this on your own footage: Ceptory indexes the video, then writes the incident report with this prompt.
See the guideHow the ai video report prompt builder works
Vague prompts get vague reports. A prompt that names the role, the sections in order, the format, and the rules gets the same report shape every time, which is what makes reports comparable across shifts and sites.
The section lists come from the report templates in Ceptory's guides, written for security, safety, logistics, construction and media teams. Turn sections off to keep a report short.
The rules matter most for footage: describe only what is visible, describe people instead of identifying them, mark anything uncertain, and attach a timestamp to every event so each line can be checked against the video.
Frequently asked questions
Which AI model can I use this prompt with?
Any model that can watch video or read a video's indexed description. In Ceptory, paste it into an index's Analyze tab and it runs against your footage.
Why does the prompt forbid guessing identities?
Identity guesses from video are unreliable and risky to put in writing. Describing clothing, colour and behaviour keeps the report factual and still lets an investigator follow a person across cameras.
Can I add my own sections?
Copy the prompt and add lines to the numbered list. Keep the rules block at the end so the model applies it to every section.
Get started
From calculators to answers
Index your footage once, then search, summarize, and get alerts across every camera.
