Product References vs Character References in AI Video

Product and character references need different acceptance checks. A product may require exact geometry, labels and controls; a character may require consistent identity, wardrobe and permitted use of a likeness. Define what must remain unchanged before generation and avoid treating a visually plausible resemblance as proof of fidelity or permission.

Documentation basis: Kuaishou announces Kling 4.0; Google DeepMind Veo. The recommendations below are editorial analysis and original teaching exercises, not a hands-on product benchmark or a forecast of search performance.

What should a product reference pack show?

Show the details that the shot must preserve, using clear permitted images. State which view is authoritative when assets differ. Keep factual labels and claims separately documented so a generated visual does not become the source of truth for the product.

What should a character reference pack establish?

Establish the authorized appearance and intended use. Confirm permission for a real person's likeness and any voice workflow separately. Do not assume that access to a photo grants permission to create a commercial performance or endorsement using that person.

How should contradictory references be handled?

Resolve differences before generation: changing packaging, wardrobe or lighting can create ambiguous instructions. Use fewer clear references rather than an uncurated folder. Explain the role of each image and remove assets that should not influence the result.

What should be checked during motion?

Inspect identity and required details throughout the action. Watch for changing logos, facial features, accessories and hand-object contact. A correct first frame does not establish consistency, especially when the subject turns or becomes partly obscured.

When should the reference strategy change?

Change it when repeated failures reveal that the required detail is not preserved reliably. Simplify motion, use a different permitted source view or replace the shot with real footage. Do not lower the factual standard merely to accept a visually attractive result.

A practical exercise

Create a reference checklist with three product details or three character-consistency requirements, plus explicit permission notes. Review a proposed shot at the beginning, midpoint and end. Reject it if any nonnegotiable detail changes, even when the overall composition is appealing.

Before you decide

• Separate fidelity from permission.
• Resolve conflicting references.
• Inspect identity throughout motion.

Sources and scope

Sources checked October 24, 2026. Follow the current documentation for your exact model, application, account and destination.

• Kuaishou announces Kling 4.0
• Google DeepMind Veo
• Creating with Runway Gen-4.5

PRODUCT REFERENCES VS CHARACTER REFERENCES IN AI VIDEO
Original Creator Lane decision worksheet | 2026-10-24

TASK
Audience / project:
Required outcome:
Current workflow and its specific limitation:
Options actually available without a new purchase:
Source material and permission:

DECISION QUESTIONS
What should a product reference pack show?
Your evidence / answer:
Unresolved detail / owner:

What should a character reference pack establish?
Your evidence / answer:
Unresolved detail / owner:

How should contradictory references be handled?
Your evidence / answer:
Unresolved detail / owner:

What should be checked during motion?
Your evidence / answer:
Unresolved detail / owner:

When should the reference strategy change?
Your evidence / answer:
Unresolved detail / owner:


PRACTICE BRIEF
Create a reference checklist with three product details or three character-consistency requirements, plus explicit permission notes. Review a proposed shot at the beginning, midpoint and end. Reject it if any nonnegotiable detail changes, even when the overall composition is appealing.

ACCEPTANCE CHECKS
[ ] Separate fidelity from permission
Evidence:
[ ] Resolve conflicting references
Evidence:
[ ] Inspect identity throughout motion
Evidence:

RESULT LOG
Option / exact version / account or interface:
Date / source URLs checked:
Input / settings / output version:
What was actually tested (leave blank if not tested):
Accepted results / failures / correction time:
Actual charges, if any (do not estimate as measured):
Privacy / permissions / delivery checks:
Decision and main limitation:
Next action / owner:
Recheck trigger:

REFERENCES
Kuaishou announces Kling 4.0: https://ir.kuaishou.com/news-releases/news-release-details/kling-ai-officially-launches-next-generation-kling-40-raising/
Google DeepMind Veo: https://deepmind.google/models/veo/
Creating with Runway Gen-4.5: https://help.runwayml.com/hc/en-us/articles/46974685288467-Creating-with-Gen-4-5