Recreating the pacing and visual language of a reference video is difficult when the only input is a vague description like “cinematic product reel.” A more reliable workflow is to treat the video as structured data: identify each shot, describe what changes, and then rebuild the sequence as prompts.
1. Start with a master prompt
Before looking at individual cuts, summarize the whole video in one paragraph. Capture the subject, visual style, lighting, color palette, camera language, pacing, and audio mood. This becomes the consistency layer shared by every shot.
A useful structure is:
- Subject: what or who appears
- Environment: location, time, weather, background
- Look: lens feel, lighting, grade, texture
- Motion: overall camera behavior and editing rhythm
- Sound: music style, ambience, dialogue, effects
2. Break the timeline into shots
Create one row per shot and record:
- start and end time;
- shot size and camera angle;
- camera movement;
- subject action;
- lighting and composition;
- transition into the next shot;
- dialogue, ambience, and sound effects.
Do not merge two shots just because they show the same subject. A cut from a wide establishing view to a close-up product detail changes the prompt and should be documented separately.
3. Separate observation from generation
Write down what is visibly present before adding creative instructions. For example:
Observation: a hand rotates a glass bottle on a dark reflective table while a narrow warm light moves across the label.
Then convert it into a generation prompt:
Slow macro push-in on a glass bottle rotating on a black reflective table; narrow warm rim light sweeps across the label; shallow depth of field; controlled luxury-product pacing.
This separation reduces invented details and makes the result easier to audit.
4. Keep continuity explicit
Repeat the details that must survive between shots: product geometry, wardrobe, character appearance, location, dominant light direction, and color grade. Vary only the camera position, action, or timing required by the storyboard.
5. Use an extraction tool for the first pass
For longer references, a tool can create the initial master prompt and shot list, which you then refine. PixMind Video to Prompt accepts a video file or direct public video URL and returns a master prompt plus a 1–100-shot storyboard with camera movement, subject action, lighting, pacing, transitions, dialogue, and sound. It also supports Excel export, which is useful for reviewing a sequence row by row.
The generated analysis is most valuable as a production worksheet, not as a single block of prose. Check the shot boundaries, remove redundant wording, and keep continuity notes consistent before using the prompts in Veo, Kling, Runway, Sora, or another video model.
A reusable shot template
Shot [number] — [duration]
Framing and angle:
Camera movement:
Subject and action:
Environment:
Lighting and color:
Transition:
Dialogue / ambience / SFX:
Continuity constraints:
A reference video becomes much easier to reproduce once it is converted from “style inspiration” into a precise sequence of observable decisions.
Top comments (0)