I.e., I'm not comparing against film, I'm comparing against what comes out of the video camera.
This particular comparison raises some interesting philosophical questions, which is why we haven't gotten to it yet. The comparison could be pretty misleading if done wrong.
Filmbox is designed to produce an accurate film look from scene-referred footage. But digital image data really has no look in any meaningful way.
We could, for example, show the video in a log color space as it’s encoded, but that's an arbitrary encoding that is not even intended for display. We could apply some "video" LUT or simulate how someone might "typically" color grade the footage, but that's a creative choice - and one that can still be performed in addition to the Filmbox emulation pipeline.
The right way to think about it is perhaps that video can be prepared to look like anything, But modern motion picture film has a fairly defined look. Filmbox is designed to provide ways of processing video that are closest to processed film. So we feel the meatiest comparison is Filmbox to actual film.
Maybe the best way to put it could be, "if someone tried to get it to look as much as film as they could, what would they lack that your tool provides"?
But in practice without a clear target and a lot of empirical data about the various properties of photochemical imaging they would end up with a subjective look that may look filmic on a limited range of shots but would not represent a dynamically functional model of the response of the photochemical process.
This is why we think the most apt comparison is our output compared with film, because that's the target look. We don't think the existing solutions do as good of a job of it as easily as Filmbox.