Make a Vertical Product Reel With Veo: Reference, Shots and Final Edit

A useful product Reel needs more than an attractive moving image. It needs a clear idea, accurate product details, a sequence that makes sense and a final export someone can watch.

Use Veo for a specific shot within that plan. Keep actual demonstrations, labels and product claims tied to evidence you can verify. This guide supplies original prompts, a storyboard and a take log. It does not include a generated Reel or claim that these prompts were tested.

Decide what the Reel should show

For an illustrative brief, imagine you own a photograph and a short real demonstration of a reusable desk organiser. The message is: “Give your everyday tools a place.” The Reel does not claim to make people more productive or prove any measured outcome.

A simple edit might contain:

This plan tells you what not to generate. If viewers need to see whether a pen fits in a slot, film that action. A model’s version could invent the slot size or movement.

Check the interface before writing a long prompt

Google documents 9:16 generation and image inputs for Veo 3.1. In the current API table, Veo 3.1 and 3.1 Fast support up to three reference images. Veo 3.1 Lite does not support the multi-reference field, although it accepts initial and final image inputs. Reference-image generation uses an eight-second duration in the supported workflow. Do not assume an app exposes every API setting. Veo 3.1 documentation

Record the displayed model, interface, aspect ratio, resolution and duration. If the tool cannot provide the controls your shot needs, simplify the shot or choose another documented workflow. Do not spend credits hoping a hidden setting will appear.

The prompts below are written as creative briefs, not ready-to-run API requests. Check current pricing and your account’s usage rules before generating.

Prepare an owned reference

Google's Ingredients to Video workflow uses supplied references to guide a scene. Use that as a reason to prepare clear assets, not as evidence that the generated product will be exact. Google's Veo reference-workflow announcement

Choose a sharp, well-lit reference that shows the actual product variant. Keep its important edges inside the frame. Remove irrelevant personal information from the surroundings.

Make a small truth sheet: shape, material, colour, parts and label. Keep the exact label artwork separately. A reference helps direct the model but does not lock every detail.

If the product must appear at a completely different angle, ask whether you can photograph that angle. A new source photo may be easier than repairing a generated view that invents the back of the object.

Generate one job per shot

A prompt that asks for a dramatic orbit, a hand demonstration, readable text and a perfect ending puts several failure points into one take. Begin with a simple camera or subject movement.

Prompt A: an environment insert

Create a vertical shot of a quiet desk corner in soft morning light. A blank notebook and a simple pencil sit on the desk, leaving a clear area in the centre. The camera makes one gentle forward movement. Keep the scene calm and physically ordinary. No lettering, logos, people or product claims. This is illustrative background footage for a desk-organisation Reel.

Prompt B: a restrained product presentation

Use the attached owned organiser image as the product reference. Show the same item standing still on a plain desk in soft side light. The camera moves slowly a small distance from left to right. Keep the complete product visible and preserve its shape and visible details as closely as possible. No hands, new compartments, written claims or added accessories. I will inspect the result against the real item.

Prompt C: a background for the ending

Create a vertical, quiet desk scene with open space in the upper half for text added later. Use soft side light and a restrained neutral palette. Keep the camera still and the background simple. No words, no visible interface and no identifiable brand. The scene should support an instructional end card rather than a dramatic reveal.

These instructions reduce complexity; they do not guarantee an accepted take.

Review a take before generating the next one

Watch at normal speed first. Does the movement help the story? Then scrub through the beginning, middle and end.

Look for changing product geometry, texture that crawls, a label that melts, an object appearing from nowhere or a shadow moving independently of the light. Inspect the entire used portion, not one attractive frame.

Mark a take with one of three decisions:

• Accept for this precise use.
• Use only a shorter clean section.
• Reject and record the reason.

Do not hide an inaccurate product demonstration by cutting quickly. Shortening is appropriate when it removes an irrelevant camera wobble, not when it disguises a factual problem.

Build the final edit outside the generator

Place the real demonstration first in the editing timeline. Build around the evidence rather than replacing it with the most visually impressive clip.

Trim the generated insert to the time the idea needs. It can be much shorter than the generated source. Match its apparent movement and colour to the surrounding footage. Add narration separately if you need exact wording. Treat any generated audio as unapproved until you have listened to it.

Keep the offer and final action editable. Proofread subtitles against the approved narration. Check the video with sound and without sound so important information is not available through only one channel.

Before publishing, inspect the product again in the final export. Cropping, speed changes and text layers can introduce new problems after a take was approved.

Record the real cost of a usable result

Do not count only the accepted generation. Your cost log should include retries, rejected takes and the time spent repairing or assembling footage.

WORK LOG — fill with actual observations
Interface and model:
Account pricing checked on:
Generation attempts:
Accepted takes:
Credits or charges shown:
Source preparation time:
Review and repair time:
Final editing time:
Total usable shots:
Final export approved by:

No cost estimate is supplied here because the actual model, plan and retries change the result. A small generation allowance is useful only if you also define a stopping rule: when the shot fails its accuracy check, return to real footage instead of continuing indefinitely.

Finish with a publication check

Confirm permissions for reference images, music and any person shown. Follow the destination platform’s current AI-media disclosure requirements. Never present a generated scene as a filmed testimonial or an actual product test.

Save the reference, prompts, accepted and rejected take notes, editable project and final export together. That folder—not one isolated AI clip—is the useful outcome of the workflow.

VEO PRODUCT REEL — PLANNING KIT
No generated video or observed test result is included.

BRIEF
Product:
Viewer:
One message:
Supported claims:
Actual proof footage:
Reference image and permission:
Actual interface / model:
Aspect ratio / resolution / duration:
Pricing checked on:
Approved maximum generation allowance:
Stopping rule:

SHOT PLAN
01 Opening: real problem footage.
02 Context: optional generated environment.
03 Proof: real product use.
04 Detail: real close-up or approved photo.
05 End: accurate image with editable next step.

PROMPT — ENVIRONMENT
Create a vertical shot of [setting] in [light]. Leave [area] clear. Use one gentle camera movement. Physically ordinary surroundings, no lettering, logos or product claims. This is illustrative background footage.

PROMPT — PRODUCT
Use the owned reference of [product]. Place it in [setting]. Keep the product still and the whole silhouette visible. Use [one small camera movement]. Preserve visible details as closely as possible. Do not add parts, labels, accessories or claims. Inspect before approval.

PROMPT — ENDING
Create a quiet vertical scene with open space at [position] for editable text. Keep the camera still, lighting simple and background low-detail. No words or interface.

TAKE LOG — repeat for each attempt
Shot ID:
Input reference:
Prompt version:
Actual controls:
Generation date:
Displayed cost or credits:
Candidate:
Accuracy failure:
Motion failure:
Audio failure:
Clean section:
Accept / trim / reject:
Reason:
Reviewer:

FINAL CHECK
Real proof retained:
Exact narration approved:
Subtitles checked:
Generated audio reviewed or replaced:
Product label and shape checked:
Rights checked:
Disclosure considered:
Final export reopened:
Project and source stored:

OBSERVED WORK
Attempts:
Accepted takes:
Charges:
Source preparation minutes:
Review/repair minutes:
Edit minutes:
These fields are blank until a real production run.