The product lesson: creators wanted scenes, frames, voice, and timing, not just generic nodes.
We originally explored a more generic workflow canvas. It made sense on paper because it looked flexible and could connect many model types. But the feedback was consistent: creators were not asking for abstract nodes. They were asking for scenes, frames, voice, motion, and timing.
That pushed us toward a storyboard-first direction. A storyboard is easier to reason about because it mirrors how people already think about short-form media: scene one, then scene two, then voice, then sound, then export.
The important lesson was not that workflows are bad. It was that the right abstraction matters. A graph editor is powerful, but if it hides the story, it increases cognitive load instead of reducing it.
By moving toward scene rails, asset reuse, motion controls, and audio lanes, the studio started feeling more like an editor and less like an internal developer tool.
That shift still leaves room for deeper orchestration later. But the product surface users touch first should match the job they are actually trying to do.