Google's third Flow Sessions cohort points its filmmaking tool at non-filmmakers

What the third cohort changed

Google Labs piloted Flow Sessions last September as a six-week program that brings artists together to build with Flow, its AI filmmaking tool. This third class marked a deliberate shift: it was the first time the team focused on recruiting creatives from outside filmmaking.

The stated reason is investigative. Google wanted to understand how Flow "might unlock something beyond traditional filmmaking," so it pulled in people from journalism, advertising and fashion with varying levels of AI experience. That framing treats the cohort less as a marketing showcase and more as a structured probe into who else might use a video-generation tool and why.

Google is explicit that the program is co-creation: "we believe the best tools are built alongside the people who use them most." In practice, that means the artists' improvisations become product feedback.

The workarounds are the real signal

The most telling detail in the post isn't any single film — it's how often artists had to reach outside Flow to get what they wanted. Julie Wieland used AI Studio to build a separate app that lowered the frame rate to fake a handmade stop-motion look, and composed her own soundtrack. Calvin Herbst trained a style transfer in a separate tool on his own archival 16mm childhood footage before bringing it into his work.

Charline Prat and the studio COMBO built reference libraries for Flow to pull from, specifically to hold visual consistency across textures and characters. That points to a known limitation of generative video: keeping a world coherent shot to shot. The artists solved it with an asset-library discipline borrowed from conventional production pipelines.

For an applied team, these are the useful data points. Each workaround marks a gap between what the tool does natively and what a serious creative project demands — frame-rate control, external style training, persistent references.

Two artists using the model's flaws on purpose

Stephane Benini's film "Echoes of Us" used what the post calls "Veo's visual drift" as a deliberate storytelling technique, alongside Flow's high volume of outputs, to convey nostalgia and impermanence. Drift — the tendency of generated frames to wander from a fixed subject — is usually a defect to suppress. Here it became a device.

Chloe Desaulles took the opposite tack. Her piece "Veneer" renders a fictional New York neighborhood that "looks remarkably real," using documentary style to interrogate realism in AI-generated media. The tool's fidelity becomes the subject rather than a selling point.

what matters is what you're trying to say before you even touch Flow.Montana Labs

What recruiting outside film tells us about Flow's roadmap

By choosing a fashion designer, a creative director-researcher and others from advertising and journalism, Google is testing whether Flow's addressable use is broader than short films. Prat's work imagined a world around a physical embroidered garment; Desaulles applied a journalistic frame. These are adjacent industries, not hobbyist experiments.

The specific implication: Google is using a curated artist program to map Flow's missing features and its cross-industry demand at the same time. The recurring pattern of external tools and reference libraries is a to-do list, and the deliberate embrace of visual drift shows the company is comfortable letting artists find value in the model's imperfections rather than waiting for them to be engineered away.