Bespoke AI Models Are Revolutionizing Filmmaking

The film industry is witnessing a significant shift with the rise of bespoke artificial intelligence models, tools specifically crafted to integrate into creative workflows. Unlike broader generative AI, these new systems are designed to be customized for individual projects, aiming to enhance the filmmaking process while navigating complex issues like copyright. This move towards customization is a key differentiator, positioning AI not as a replacement for artists but as a specialized toolkit that adapts to their unique vision.

A prime example is Netflix’s recent acquisition of the AI startup InterPositive. The streaming giant highlighted the technology’s customizability as a major factor, framing it as a way to empower filmmakers. While financial details were not officially released, reports suggest the deal could be valued around $600 million. Netflix plans to use InterPositive’s core technology by training it on “dailies”, the raw footage from ongoing productions. This creates a project-specific model that directors and cinematographers can later use in post-production for tasks like adjusting lighting, removing rigging equipment, or seamlessly altering backgrounds. The central idea is that because the model learns directly from the project’s own footage, its outputs will naturally align with the established creative aesthetic.

InterPositive’s founder, Ben Affleck, described building the system by filming a proprietary dataset in a controlled studio environment. The goal was to create a tool that speaks the language of filmmakers, focusing on techniques like cinematography rather than replicating performances. This focus on controlled, licensed datasets is presented as an ethical alternative to models trained on vast, unlicensed swaths of internet data. The approach aims to provide consistency and creative control, though its effectiveness relies on the core model being trained on a sufficiently diverse range of production scenarios to handle any directorial request.

This model is similar to the strategy of another studio, Asteria. Its flagship product is also a proprietary AI model that clients can customize with their own original art. Asteria recently announced an AI-powered operating system that can analyze a script and automatically generate detailed databases covering characters, scenes, and budgets. While InterPositive seems geared toward fine-tuning existing footage, Asteria’s technology has been used to generate entirely new assets, like characters and background objects, that share a cohesive style drawn from the licensed dataset.

For producers, the appeal is clear: the promise of faster, more cost-effective production timelines. The ability to generate consistent visual elements or make rapid post-production changes without extensive reshoots can significantly impact a project’s bottom line. This economic incentive is a powerful driver for traditional studios to adopt these technologies. The industry’s direction is further signaled by partnerships like Adobe’s collaboration with major studios to develop “IP-safe” models for its creative software suite.

However, the long-term impact on the creative workforce remains an open question. While companies like InterPositive and Asteria emphasize empowering artists, the tangible benefits for those artists are less defined. Streamlining production can increase studio profits, but it does not automatically translate to more job security, higher wages, or better working conditions for human creatives. The narrative focuses on efficiency and creative assistance, yet the details on how these tools will genuinely augment rather than simply accelerate human labor are often vague. This gap between promise and practice warrants a measured perspective as these bespoke AI models become more entrenched in the filmmaking process.

(Source: The Verge)