Flow AI Video Prompts: Camera, Motion and Scene Examples
On this page
Quick Answer
A useful Flow AI video prompt describes one scene, one primary action, one camera behavior, and one visual result. Add timing, aspect ratio, lighting, and audio only when they matter to the final clip.
Draft the brief here, then test it in Flowith AI Video without changing every variable between models.
Flow AI Video Prompt Formula
Use this order:
subject + setting + action + camera + lighting/style + timing + aspect ratio + audio + exclusions
Example:
A ceramic coffee cup on a walnut desk at sunrise. Steam curls upward while the camera makes a slow six-second push-in. Warm side light, shallow depth of field, realistic product photography, 16:9, no text or logo changes.
The structure matters more than decorative adjectives. If the subject, action, and camera instruction conflict, the model has to guess which one to follow.
Prompt Examples by Job
| Job | Prompt direction | Review first |
|---|---|---|
| Product shot | Keep the product shape fixed; specify one camera move and one lighting change | Logo, proportions, reflections, text |
| Social clip | Put the action in the first second; use vertical framing and one readable focal point | Opening frame, crop, pacing |
| Cinematic scene | Name the subject path, lens behavior, atmosphere, and shot duration | Motion continuity, camera drift |
| Character shot | Keep clothing, face, hair, and environment consistent; limit simultaneous actions | Identity, hands, background changes |
| Image animation | Describe what moves and what stays fixed | Frame preservation, unwanted morphing |
Camera Instructions
Use one primary camera behavior per first generation:
- static shot for product accuracy or subtle subject motion;
- slow push-in to build attention without changing the scene;
- pull-back reveal when the environment is part of the story;
- left-to-right tracking for movement across a space;
- gentle handheld for documentary energy;
- locked overhead shot for tutorials, food, or tabletop work.
Avoid combining orbit, zoom, pan, crane, and handheld movement in one short clip. Test the most important movement first.
Motion Instructions
Separate subject motion from camera motion:
- Subject: “The runner crosses from left to right and slows at the doorway.”
- Camera: “The camera tracks beside the runner at waist height.”
- Fixed elements: “The storefront sign and street layout remain unchanged.”
For a still reference frame, use the Flow AI image-to-video guide before adding complex motion.
Audio and Dialogue
Only request audio when the selected model and live generation mode support it. Keep dialogue short, specify the speaker, and review pronunciation, synchronization, background sound, music rights, and disclosure before publishing.
If audio is not essential to the visual test, validate the shot first and add sound in the editing workflow.
How to Compare Models Fairly
- Keep the scene brief and reference frame unchanged.
- Use the same aspect ratio and closest available duration.
- Record the selected model and generation mode.
- Score subject consistency, motion, camera fit, artifacts, audio, and usable seconds.
- Change one prompt variable before the next round.
Use the AI video workflow, pricing, and model guide when revision cost or provider choice becomes the main decision.
Frequently Asked Questions
How long should a Flow AI video prompt be?
Long enough to define the scene and constraints, but short enough that priorities remain clear. Start with two or three precise sentences rather than a page of competing instructions.
Should I name a camera lens?
Only when lens behavior matters. A plain instruction such as “wide establishing shot” or “close product shot with shallow depth of field” is often easier to review.
Why does the subject change during motion?
The prompt may ask for too many simultaneous changes, or the reference frame may not define the subject clearly. Reduce motion and state which details must remain fixed.
Can one prompt work across every model?
Use one shared brief for comparison, then adapt only the syntax or controls required by the selected model. Do not assume identical support for duration, audio, references, or editing.
Next Steps
- Start the generation in Flowith AI Video.
- Prepare a still with the Flowith AI Image Generator.
- Check Google Flow access, credits, and Veo modes.
- Compare model constraints in the AI video workflow guide.