The production work behind Grimes’ AI-assisted “Accelerando”

Grimes’ representative told Business Insider on 13 October that no AI was used to record “Accelerando” or its vocals. Generated visual elements were incorporated into a wider effects process involving more than 70 people over two to three months. Live action was filmed against a blue screen, with real performers and physical elements.[1]

The credits, reproduced by MDX Music alongside the official video, put names to that work. Cho Giseok directed; Grimes was creative director; ORB HAUS handled production and post-production/VFX. They also name choreographer Hyoseung Ye, dancers, camera and lighting crews, prosthetic makeup specialists and a sculptor. The production extends well beyond the generation of images.[2]

The jobs between an image and a shot

Two of the less familiar credits are especially useful for understanding the process. The list names 75MM Studio for matchmove and a team of compositors. It credits Zaddicted as VFX supervisor and lead, alongside other effects and 3D artists. Each describes a different responsibility in constructing the final picture.[2]

Matchmove makes computer-generated material share the camera’s movement and perspective with filmed footage. ScreenSkills describes the job as recreating the live-action camera digitally, including properties such as lens distortion, so the elements can line up. As an illustrative example, a virtual structure should stay planted in the scene when a camera moves past a performer. Its apparent position cannot drift independently of the ground.[3]

Compositing brings the separate materials into a coherent image. The work can combine live footage, computer graphics and painted backgrounds, then reconcile colour and lighting. ScreenSkills also describes keying: extracting a performer from a distinctive blue or green background so they can appear in another setting. That separation is only one part of making the resulting shot hold together.[4]

A generated element could therefore be an input to a much longer sequence of decisions. Someone still has to choose where it belongs, how large it appears, how it moves relative to the camera and how its light relates to the people in the frame. This is an explanation of the credited crafts, rather than a reconstruction of which specific shots used AI.[3][4]

Credit at the level of the contribution

The account names neither the models used nor specific generated shots, so it cannot quantify AI’s visual contribution.[1]

What the credits do provide is a more useful way to discuss the work. Music recording, performance, image generation and the assembly of a moving scene are separate contributions. Naming the people responsible for them lets a viewer ask a more exact question about an AI-assisted film: which part did the tool supply, and what did the rest of the production make from it?

Report an error or suggest a correction