AI video generation can transform an ordinary portrait into a convincing moving scene. That creative power creates a strict responsibility: the person in the source image must have the right to decide whether and how their likeness is transformed.
1. Establish ownership and consent
Use media you created yourself, licensed stock that permits generative transformation, or an image supplied by a consenting adult for a clearly defined purpose. Save the license or written permission with the project record.
Consent must be specific. Permission to photograph someone does not automatically authorize animation, intimate transformation, advertising, public posting, or commercial use. If the intended output changes, ask again.
Never upload:
- any image of a minor or a person whose age is uncertain;
- intimate or private media;
- a former partner, acquaintance, coworker, or stranger without explicit permission;
- identification documents, medical images, private messages, or location evidence;
- celebrity images for deceptive endorsement or impersonation.
2. Minimize source data
Crop out bystanders, addresses, screens, badges, documents, and unique location details. Remove EXIF metadata when it is not needed. Use the smallest resolution that still allows a meaningful quality test.
Keep the source file in controlled storage and do not treat a cloud generation history as your only archive. If the image is sensitive enough that exposure would harm the subject, do not upload it to a generative service.
3. Read the current product rules
Before generating, inspect the current official terms and privacy page. Look for:
- how uploads and outputs are stored;
- whether material may be reviewed by people;
- whether content may be used to improve models;
- how to delete an upload, output, collection, or account;
- whether public sharing is automatic or optional;
- how content removal and rights complaints work;
- restrictions on commercial publication.
Policies can change. Save the date and link used for an important project rather than relying on memory or an old third-party review.
4. Generate privately and review slowly
Keep early tests private. Watch each clip frame by frame and reject outputs that distort identity, imply actions the subject did not approve, expose private details, or could reasonably mislead an audience.
Do not publish an output simply because the model produced it. The creator remains responsible for context, consent, disclosure, and downstream harm.
5. Label synthetic media
If a reasonable viewer could believe the footage records a real event, clearly disclose that it is generated or altered. Preserve available provenance information and keep a production log with the source, permission, prompt, date, model, and edits.
Disclosure is especially important for advertising, testimonials, news-like presentation, and realistic depictions of a recognizable person. It does not make an unauthorized use acceptable.
6. Delete and verify
After testing, remove unused uploads and outputs through the available controls. Check collections, reels, trash, history, and account settings separately. If a removal route is unclear, use the official support or content-removal channel and keep a record of the request.
Bottom line
Musebox AI safety begins before the upload. Use only rights-cleared adult media, minimize personal data, understand current retention and deletion terms, keep experiments private, disclose realistic synthetic footage, and stop immediately if consent is missing or ambiguous.