Responsible AI
Use & Disclosure
Marley Digital Media, Inc. makes stories with AI as a tool in human hands, and tells you plainly when it does.
Last updated November 17, 2026 · Ask a question or send feedback
Our promise
Every story we release tells you how it was made. These are the commitments behind that:
- People lead. A human writes, directs, edits and approves every MDM release. AI is a tool in that work, never the author of it.
- We disclose AI use. Any release that contains AI-generated picture, sound or voice says so, in the credits and in the description.
- The watermark stays in. We keep the invisible AI watermarks our tools embed. We never strip, alter or hide them.
- No fake people. We do not make AI likenesses or voice clones of real people without their written consent.
- Culture is not a costume. Stories drawn from living traditions are shaped with people from those traditions, and AI never imitates a community's accent or language in their place.
- We credit the humans. Performers, writers, consultants and musicians are named for their work.
- We answer questions. If you want to know how something was made, ask us and we will tell you.
How we use AI
MDM is an AI-native studio: we use generative tools to build worlds an independent studio could not otherwise afford to make. The creative decisions stay with people.
| Stage | What people do | Where AI helps |
|---|---|---|
| Story | Write the series bibles, scripts and canon rulings | Research, drafting support, continuity checks |
| Design | Approve every character, costume and set | Concept images, character sheets, set plates |
| Picture | Direct each shot, choose the takes, cut the film | Video generation for shots and pickups |
| Sound | Score, mix and master; cast the final voices | Temporary voices for drafts, sound design |
| Finishing | Grade, title and approve the release | Upscaling and technical clean-up |
Our main tools include Google Flow and Veo, Higgsfield, Runway and MDM's own music models, finished in DaVinci Resolve, Premiere, Logic Pro and Blender. MDM's music model is trained on releases the company owns.
How you'll know
Every MDM release carries AI disclosure in four layers, so it survives wherever the work travels.
| Layer | What it is | Where you find it |
|---|---|---|
| Invisible watermark | Google's SynthID, embedded in every frame our Google tools generate. We keep it in and never remove it. | Inside the file itself, readable by verification tools |
| Content credentials | Machine-readable provenance data that some of our tools attach (C2PA) | In the file's metadata |
| Credits and description | A plain-language AI disclosure line on every release | End credits, video description, release page |
| Platform labels | The AI-content label on platforms that offer one, such as YouTube, for realistic AI imagery | Below or on the video |
We turn off visible corner watermarks so they don't interrupt the story. The disclosure moves to the credits and labels instead; it never disappears.
Collectibles and releases on MDM's Sovereign Mint also carry onchain provenance: a SHA-256 fingerprint of the file, stored on IPFS, so anyone can confirm a piece is the original.
Sample credit line:
Made by Marley Digital Media, Inc. with AI-assisted picture and sound. Story, direction, edit and final approval by people. AI-generated imagery carries Google SynthID.
Voices, performers and likeness
The voices that reach you in a final release belong to people we cast and credit.
- Temporary voices stay temporary. In drafts and early teasers we may use AI voices as placeholders, labeled "temp" in the credits. They are replaced by human performers for the final work.
- No imitation accents. An AI voice never imitates a real community's accent or language. Where a role calls for a Ghanaian, Akan or other cultural voice, we cast a performer from that community.
- Consent for every likeness. We create a person's voice clone or digital likeness only with their written consent, for the uses they agreed to, and we credit them.
Culture and community
AF3 stories draw on Black history and living African and diaspora traditions, so we hold them to a standard set by people, not by software.
- The Griot Protocol. Every AF3 property keeps a canon log where each creative decision is recorded, sourced and ruled on by a person, so the story's lineage stays traceable.
- Traditional stories stay distinct. We mark what comes from tradition (for example the Akan Ananse stories) and what MDM invented, so neither is mistaken for the other.
- Cultural consultation. Portrayal, language, names, symbols such as Adinkra, and pronunciation are reviewed with people from the culture before a release is final.
- Melanin-true imagery. We grade and review Black skin tones for true depth, warmth and texture, and reject AI output that lightens, grays or distorts them.
Tell us what you think.
If you want to know how a piece was made, or think we got something wrong, use this form. We read every message and update this page as our tools and the rules around them change.
