An impact video can imply a beneficiary story that never happened. A generated family receives clean water, a worker smiles beside a new facility, or a child enters a bright classroom. The narration may use aggregate facts, but the images encourage viewers to read them as documented outcomes for real people.
CSR communication needs a person-and-claim provenance matrix. Every person shown is classified as documented participant, consented reenactment, licensed stock, or generated illustration. Every outcome claim remains linked to its own evidence. The two records meet in the edit, but they never substitute for each other.
Separate the Human Story from Outcome Evidence
List the claims the video intends to make: funds delivered, infrastructure completed, people reached, behavior changed, or conditions improved. Record source, period, geography, sample or counting method, owner, and uncertainty. Then list every person or community image separately.
Distinguish outputs from outcomes. Building a well, distributing kits, or holding training can be documented as activity. Improved health, income, attendance, or resilience needs separate evidence and time. Do not let a visual sequence turn completion of an activity into proof of long-term change.
Record the attribution boundary where relevant. A project can contribute to a change without causing all of it. Keep verbs such as supported, reached, installed, or reported aligned with the source instead of upgrading them to transformed or solved.
Classify People before Choosing Emotional Scenes
A documented participant appears in verified project material with appropriate consent. A reenactment is performed and labeled. Licensed stock illustrates a category. A generated person is synthetic. None should be presented as another category because the scene feels representative.
Consent is specific to use, audience, duration, and context. A photograph approved for internal reporting may not be approved for a public campaign. Record the consent scope rather than treating possession of an image as permission.
Plan a withdrawal route. Participants may need contact information, safety protection, or the ability to revoke future use under the applicable agreement. The asset record should show where each version was distributed so a decision can reach derivatives.
For children or vulnerable people, apply the organization’s safeguarding process before editorial work. Do not assume that a guardian signature resolves dignity, future searchability, or contextual harm.
Bind Every Outcome to Its Evidence Record
Give each claim an identifier that follows narration, captions, charts, and source notes. Aggregate evidence should remain aggregate. Do not place one generated face beside a statistic in a way that implies the person contributed to or received the measured outcome.
Map time and geography too. A national result beside a village scene can imply that the place produced the statistic. A current portrait beside a five-year aggregate can imply a relationship that the evidence does not establish.
Keep uncertainty, sample size, and evaluation method near outcome figures. The video may simplify presentation, but it should not remove the condition that makes the number interpretable. Use an accompanying source page when the full method cannot fit.
| Visual subject | Permitted role | Required label or record |
| Documented participant | Specific project story | Consent and factual verification |
| Reenactment | Explain a process | Visible reenactment label |
| Licensed stock | General context | License and non-specific caption |
| Generated person | Conceptual illustration | Illustration disclosure |
Compare Documentary and Illustrative Video Roles
Documentary material can support a specific account when its provenance, consent, and context are known. Illustrative video can explain a mechanism or provide non-specific atmosphere. It cannot establish that a project reached a depicted person, place, or result.
Use Documentary Material for Specific Outcomes
Verify names, roles, dates, locations, quotes, and before-and-after relationships. Avoid staging a participant to repeat a claim they cannot verify. Keep original media and interview records with the edit so a reviewer can trace the account.
Review power and dignity, not only legal permission. Remove unnecessary health, financial, identity, or location details. A moving story should not make a participant permanently searchable for a condition they disclosed in a specific context.
Show agency rather than staged gratitude. Let participants describe choices and work where the record supports it. Avoid editing that frames people only as recipients while the organization appears as the sole actor.
Check whether before-and-after images share subject, date range, viewpoint, and conditions. If they do not, treat them as contextual images rather than a visual measurement of impact.
Use Generated People Only as Declared Illustration
MakeShot.ai can route text or reference images into video candidates. An AI Video Generator may create a conceptual scene when no real person is being documented. Use non-identifiable settings and avoid agency uniforms, project signage, exact facilities, or demographic shorthand for eligibility and need.
Keep the illustration label close enough to survive crops and syndication. The narration should say “illustration” or otherwise avoid a documentary construction. A footer that disappears on social video does not preserve provenance.
Use fictional, non-diagnostic actions such as carrying generic materials or viewing an abstract plan. Do not show cash transfers, medical treatment, certificates, or completed infrastructure unless the scene is explicitly conceptual and cannot be confused with the reported project.
Inspect generated signs, uniforms, tools, and architecture. Plausible details can falsely identify an agency, country, occupation, or facility. Simplify or remove them before the candidate reaches the edit.
Compare the Implied Story without Narration
Show the picture track alone and ask who viewers believe the people are, what happened to them, and whether the scene looks recorded. If reviewers infer a real beneficiary or completed outcome, change the imagery or strengthen the disclosure.
Then listen to narration alone. Check that every outcome remains bounded by its evidence and that no singular story has been invented from aggregate data. Picture and sound should converge without creating a third unsupported claim.
Ask the reviewer which scenes they believe are documentary, reenacted, stock, or generated. Any confident misclassification is a disclosure failure even if a legal footer is present. Make the production role understandable at normal viewing speed.
Review Dignity Consent and Claim Proximity
Inspect how close each person appears to each statistic, quote, or brand assertion. Physical proximity in a frame implies relationship. Separate conceptual people from specific result cards unless the record supports that relationship.
Reject Every Stereotype Used as Visual Shorthand
Do not use clothing, housing, disability, age, skin tone, or family structure as a shorthand for poverty, need, gratitude, or program success. Describe the setting and action required by the explanation, not a demographic profile expected to trigger emotion.
Review translations and local context. A gesture, occupation, or environment can carry different implications across audiences. Local review should be recorded as editorial evidence, not used to claim universal interpretation.
Compare who speaks, who acts, and who is watched. If staff receive names and agency while community members remain silent background figures, the edit may reproduce a power imbalance even without an explicit false claim.
Audit music and pacing. Sentimental scoring or slow-motion close-ups can turn ordinary activity into a plea or celebration not supported by participants’ accounts. Emotional tone belongs in the dignity review.
Keep Generated Faces out of Testimonials
A generated person must not mouth a real participant’s quote or appear beside quotation marks as its speaker. Keep verified wording on a separate quote card with the real speaker’s authorized attribution, or paraphrase aggregate findings without constructing a testimonial.
A second AI Video Generator candidate is justified when disclosure, setting, or stereotypes fail review. It cannot repair missing outcome evidence or consent.
Release the Video with a Provenance Matrix
Package claim sources, person classifications, consent records, MakeShot.ai prompts and candidates, labels, review notes, and final exports. Read back disclosure at every delivery size and keep a correction owner.
MakeShot.ai can provide conceptual motion, but only the matrix prevents illustration from becoming a convenient fake case study. Publish when every person has an honest role and every outcome can stand without borrowing credibility from a face.
