Today we completed the production phase of a new Directed Generative Production test, exploring how live performance and structured scene information can guide generative image-making.

The methodology starts with a principle that has become fundamental to how we think about generative production at Wander•:

Generative AI is the renderer, not the author. Production data, not prompts, drives the image.

For us, the interesting question isn’t simply whether generative technology can create an image. It’s whether it can become part of a genuinely directable filmmaking process.

Can a filmmaker ask to change one specific element without losing the performance, composition or creative decisions they’ve already approved? Can generative finishing behave less like a new roll of the dice and more like an established production tool that can respond to direction and revision?

That’s what we’re testing.

For the production phase, filmed performance was combined with structured scene information, motion-capture data and established VFX methodologies to create the foundation for the generative work that follows.

Just as importantly, introducing generative technology didn’t mean removing the people traditionally involved in making the work. The shoot brought together an experienced production crew, with performance, photography and creative direction at the centre of the process.

We believe that’s an important part of the conversation around where this technology goes next.

The opportunity we see isn’t in replacing established filmmaking with generative production. It’s in determining where generative technology can extend what filmmakers and artists are already capable of doing, while preserving authorship, directorial intent and the ability to make deliberate creative decisions throughout the process.

With the production phase now behind us, we’re moving into generative finishing and evaluation. That’s where we’ll begin putting the methodology through its real test: making changes, giving notes, preserving approvals and finding out where the approach works, where it doesn’t and where it can meaningfully improve the filmmaking process.

There is still plenty to learn.

That’s precisely why we’re testing it.

Update, 30 September 2026: Post-production is now nearing completion. Read our latest development update.