AI Film Infrastructure Is Forming in Public

A production desk where the next shot waits for approval

Two announcements this month say the same thing from different ends of the pipeline. Frame Rate Oy is shipping a production layer that sits above the video models. COL Group is standing up an AI studio inside a distribution network that already reaches thousands of titles. The waitlists open in weeks. The claim on the studio side is hundreds of AI productions already delivered.

If you write, direct, or teach, this is a good moment to look at how the work is being organized, not only which checkpoint rendered the last clip.

footagen.ai, above the models

Finnish company Frame Rate Oy unveiled footagen.ai, an AI-native film workflow that starts from a screenplay and builds structured scenes, shots, characters, locations, and props before generation begins. From there it moves into a node-based board where you can pick a model shot by shot and keep a human approval gate in the loop. The models they cite today include Seedance 2.5, Hailuo H3, Nano Banana 2, GPT Image 2, and Seedream 5 Pro.

Storage is local-first by default, and Content Credentials via C2PA are on the roadmap. The waitlist opens on September 10, 2026. The public debut is at IBC2026 in Amsterdam, September 11 through 14, on the EIT stand FT1407.

CEO Jukka Karhula put the design intent in plain language: AI should not take the director’s chair, and the product is not built to mass-produce disposable footage. That is the right bar for a tool that wants a place on a real schedule. A screenplay that becomes a shot list, then a set of model choices under human sign-off, is something you can teach, audit, and improve.

COL AI Studios, at distribution scale

Variety reports that COL Group has launched COL AI Studios to help platforms and brands test and scale IP across animation, short and long form, branded entertainment, and live action. The parent network lists more than 5,000 live-action titles and more than 500 creatives across Los Angeles, China, and Southeast Asia.

The studio says it has already delivered more than 300 AI productions. An early franchise, Love Missions Across the Galaxy, runs three seasons and more than ten languages. GM Timothy Oh describes AI as another way of delivering a story, and adds that greater access does not automatically lead to better entertainment. That second half is the useful half. Access is arriving. Taste and craft still decide what survives a season.

Why the speed matters

Put the two stories next to each other and the shape of 2026 becomes clearer. One company is building the production desk that sits on top of Seedance, Hailuo, Seedream, and the rest. Another is putting AI development inside a distribution machine that already knows how to move titles across markets and languages. Waitlists open in a matter of weeks. Franchise seasons are already shipping. The infrastructure is forming in public, with named stands, named executives, and named early titles.

For anyone learning this craft, the homework is concrete. Walk the footagen waitlist announcement and the IBC booth notes. Read the COL launch against Oh’s caution about access and quality. Sketch how one of your own scenes would look as structured shots with a human approval gate. The models will keep changing. The people who learn to organize the work around them will be the ones who keep making pictures audiences want to finish.

Time to dig deeper

Start with Frame Rate’s IBC writeup and Variety on COL AI Studios. Then open footagen.ai, note the September 10 waitlist and the Amsterdam dates, and ask yourself where your own next project would put the human approval step. That question will age better than any single model name on the node board.