Extracting a prompt from a video means translating visible evidence into new production instructions. rea-video breaks down shots, action, captions and sound so you can draft per-shot prompts and a production plan. It cannot recover the exact original prompt or generate a video on this website.
Start with one real shot
Shot 25 of the unboxing case runs from 00:22.900 to 00:23.567. The report describes the left hand holding a white headphone box while the right hand slits the seal with a small yellow box cutter, filmed from directly overhead with a fixed camera. Separate those observations before drafting a prompt.
| Field | Description from the report |
|---|---|
| Subject | Hands, white headphone box, small yellow box cutter |
| Action | The left hand holds the box; the right hand slits the seal |
| Camera | Top-down medium view, fixed camera; subject moves |
| Post-production | Handle text and graphic overlays separately; check timing in the report |
The report classifies this shot as live action (78%) and recommends reshooting it; a generation prompt is only a fallback. The draft below has not been validated through generation.
Top-down medium view of hands holding a white headphone box. The right hand slits the seal with a small yellow box cutter. Fixed camera, stationary background, soft warm indoor light. Add text and graphics in editing. Test the action, then trim to the target length.
Draft a per-shot prompt yourself
This builder runs in your browser and formats what you observed into the structure real reports use. It does not generate video, upload anything, or recover the original prompt.
Prompt draft
See the unboxing case · Check the full report
How to generate prompts from a reference video
- Prepare a local video you are entitled to analyze and run rea-video in your agent. Check cut points and timecodes first.
- For each shot, record subject, start state, action, end state and camera movement. Separate observation from interpretation.
- Separate filming, generation, screen recording and code reconstruction. Plan captions, music and transitions independently.
- Test one representative shot. Compare composition, motion and timing before extending the workflow to the full sequence.
Installation and first run · Download the shot list template
How to get the prompt from a video (and what you actually get)
A finished video cannot reliably reveal its original prompt, seed, model version or editing project. Similar images can come from different inputs, and a clip may combine filming, generation and editing. Treat extracted prompts as a new evidence-based production plan. For exact inputs, ask the creator for the project or generation records.
Free tools, cloud models and local processing
The rea-video core is open source, but your AI assistant or model provider may charge separately. Local processing is not a guarantee that media stays on your device: frames, audio or video may be sent to a cloud model depending on configuration. Compare input requirements, costs, data handling, per-shot output and export options.
Continue with your next task
- YouTube वीडियो से प्रॉम्प्ट: रेफ़रेंस का प्रति-शॉट विश्लेषण
- AI से वीडियो का विवरण कैसे लिखें: सारांश से आगे
- कैमरा मूवमेंट प्रॉम्प्ट: कैमरा और एक्शन को अलग पहचानें
- आपका वीडियो प्रॉम्प्ट रेफ़रेंस से मिलता क्यों नहीं
- रेफ़रेंस से वीडियो प्रॉम्प्ट कैसे रिवर्स-इंजीनियर करें
- 2026 के सर्वश्रेष्ठ वीडियो-से-प्रॉम्प्ट टूल्स: ईमानदार तुलना
Adapt a draft for your model
These guides offer report-based planning examples, not verified model recreations. Check your model’s supported inputs and settings before use.