Four words carry the whole model. Once they are straight, most of the software stops needing to be explained.
| Word | What it is |
|---|---|
| Lot | The project, a named shot list with a look, a brand and an aspect |
| Shot | One scene in it: a prompt, a duration, a model, its references |
| Take | One render of one shot. A shot can hold many |
| Mounted | The take that is the shot’s answer, what export and the review link show |
A lot is a list, and that is the point
The obvious way to build this product would be to generate a finished video. That would make every complaint about it a complete re-run: one shot is too dark, so pay for the whole thing again and hope the other five come back the same.
A lot is a list of shots that each render independently. One note from a client costs one shot. A model that turned out to be wrong for the product costs the shots it made, not the campaign. This is the single structural decision the rest of the software follows from.
Takes accumulate, nothing paid for is thrown away
Render a shot four times and the shot holds four takes. They all stay. Re-rendering does not overwrite; it adds. You paid for each of those renders, and a product that deleted one because a newer one arrived would be spending your money and then discarding the result.
So the question is never which take survived. It is which one is mounted, and that is a decision you can change at any time and change back.
Mounting: the rule that surprises people
The first successful take of a shot mounts itself.
That is deliberate, it is what makes a one-take flow work, so a lot rendered once is immediately a lot you can export. But it means the take on show is the first one that finished, not the one you would have chosen.
When you render several variants of a shot that has no winner yet, the one that happens to come back first becomes the shot’s answer. The render tells you when that is about to happen, so it is a warning rather than a discovery. Overrule it whenever you like: picking a different take changes what the export ships and what a client sees, and the others stay exactly where they are.
Labelling, which is not decoration
A shot with four takes cannot tell them apart by the shot’s own title. Naming a take fixes that, and the export names its files after it: the take’s label first, then the shot’s title, then the bare number.
So labelling is how a delivered folder becomes readable at the other end. It is editorial only: it reaches no provider, changes no price, and clearing it puts the name back to the shot’s.
What a lot carries besides its shots
Three things sit above the shot list and apply to all of it: the look (the grade, genre, film stock, lighting, palette, camera language), the brand, and the aspect. The look is the one worth knowing about, because of when it applies.
It is never written into a shot’s prompt. It is appended at render time, which is why a shot reads unchanged after you set one, and why changing the look does not reach shots that have already rendered. Their takes were composed under the old one. Set the look before the first render; change it halfway and every mounted take now disagrees with the shots rendered after it.
Two shots are deliberately outside all of this. A spoken shot gets no look at all, its prompt is the script, and a model that scores its own audio would read the grade aloud. A still gets everything except the camera move, because a photograph has no motion to give.
Archive, rather than delete
This is the only irreversible thing in the whole model, so it is worth stating plainly.
Archiving a lot moves it out of the default listing and destroys nothing. Deleting one destroys the shot graph, and the shot graph is the record of what was billed. The rendered files survive; the account of what was made and what it cost does not. A workspace that deletes its lots keeps the clips and loses the ledger.
Almost every time somebody means “get this off my screen”, archive is the answer.