"Clandestina" is Emma Peters' French cover of a song about a woman adrift – her family destroyed by the drug trade, searching for connection in the wreckage. It's one of our favorite songs. We wanted to see if we could bring it to life.
This is our first fully AI-generated music video. Every frame began as a still in Midjourney, then moved through Veo, Kling, Runway, Minimax, Seedance, and Wan – whichever model gave us what we needed. Color, lens flares, and aberrations were handled in post. The result is a proof-of-concept for a new way of working: a video that would have cost $3 million to produce traditionally, completed in three weeks.
Clandestina (Cover) Emma Peters & Edmofo Remix
Composer: Lartiste
Role: Concept. Design. Direction. Production.
"Clandestina" is Emma Peters' haunting French cover of a track that tells a story of loss, displacement, and the human cost of the drug trade. A woman – a clandestina – has lost everything. Her family destroyed by fire, by violence, by the chain of addiction that runs from South America to the United States. "So that some Gringos can have their fix, lives get sacrificed." She's left with nothing but a plea: love me, take me in your arms, I've got nobody left.
It's one of our favorite songs. We'd wanted to visualize it for years.
This is our first fully AI-generated music video – a proof-of-concept for what's now possible. We set the story in Miami's neon edges: a woman navigating the city alone, memories of her family flickering through, the weight of what she's lost visible in every frame.
Every shot began as a still keyframe generated in Midjourney. From there, we ran images through every video model available – Veo, Kling, Runway, Minimax, Seedance, Wan – selecting whichever output best served the moment. Some tools handled motion better. Some handled faces. Some gave us texture we couldn't get elsewhere. Color grading, lens flares, and chromatic aberrations were all done in post.
The tools evolved as we worked. New models, new updates, sometimes weekly. We adapted in real time.
The Point
This video would have cost roughly $3 million to produce traditionally – locations, talent, crew, equipment, time. We made it in three weeks.
No single AI model does everything well. VEO handled certain motion. Kling gave us texture. Hailuo nailed faces. Runway, Seedance, Wan – each had a strength. We ran every shot through whichever tool served it best, often testing the same keyframe across multiple platforms before choosing. Below, some of the R&D and scenes that didn't make the cut – the experiments that led to the ones that did.
Copyright © MMIV-MMXXVI METAphrenie Inc. All rights reserved. Images from this site may not be reproduced without prior written permission. Designated trademarks and brands are the property of their respective owners.
