
Every sample on the site carries its own four settings. Open a finished project and you can see the choices playing out across the character sheets, the location plates and the shots.
This is the medium and the surface: photographic, painterly, cel-shaded, 3D. It reaches everything — character sheets, location plates, item shots and the video prompt itself — because it is the one layer that has to stay identical in every frame of every episode.
It does not give framing or composition instructions. Telling the system how to paint is fine; telling it where to put the subject is the storyboard's job, and the two fight if both try.
Space, materials, colour discipline, how densely the set is dressed. This one is injected in exactly one place: the script read. The model digests it once, then writes it into every location description and every shot, so the courtyard and the corridor come from the same design language rather than being re-styled independently.
It is deliberately kept out of the location images and the video prompt. Pasting it in a second time is how you get red lacquer columns growing in an abandoned courtyard — the description already said what the space is; repeating the label overrides it.
Because it lives at the read, changing it later means re-reading the script.
Garment construction: silhouette, cut, fabric, craft. Like art direction, it is injected only at the read, and for a second reason: clothing is already described per character. Adding a project-wide wardrobe paragraph on top flattens the class gradient the read worked out — the servant and the magistrate end up dressed with equal expense.
It also protects a fix we made earlier: if wardrobe is pushed into the character sheet at full strength, a long coat eats seventy percent of a vertical frame and the face shrinks to nothing.
Short drama, long-form or film. It is not a look at all: it sets the per-episode duration ceiling, the performance grammar and the dialogue rhythm. A short drama needs a hook at the end of every episode; a film does not, and the read and the storyboard both branch on this.
Paste the script and the settings are pre-filled for you — a small, free pass over your text picks the three visual axes and says why. It only fills in axes you have not touched yourself, and you can override any of them before the read starts.
All three visual axes have to be set before the read can begin. That is on purpose: they are consumed at the read, so a blank one is a decision made by default rather than by you.
| Setting | Controls | Changeable later? |
|---|---|---|
| Picture style | Rendering medium and surface | Yes — regenerate images |
| Art direction | What the set is built from | Needs a re-read |
| Wardrobe | Garment construction | Needs a re-read |
| Content type | Duration cap and pacing | Needs a re-read |
Because it is consumed once, at the read, and written into every location description from there. Changing the label afterwards would leave the descriptions saying something else.
Members can. Presets cover the common ground; a custom line is useful when you have a specific reference in mind.
Regenerated images will use the new style, and ones you already generated will not. Location plates exist so that the look holds across episodes, so switching mid-project means regenerating the sheets you want to keep aligned.
No, and that is deliberate. Light position, atmosphere and grading are already decided per shot by the storyboard. A project-wide mood setting would fight every shot that asks for something specific.
Paste your script and take the pre-filled suggestion. It reads your text and explains each pick in one line.
Open a finished sample and look at how one set of choices carries across a whole episode.
Updated September 21, 2026
Pricing·Novel to Video·Guides·Samples·Support·Privacy·Terms·Legal·Contact·© 2026 SceneMixer