AI Features
Detect production elements with AI
AI element detection reads your screenplay scene by scene and identifies the production elements in it — cast, props, vehicles, wardrobe, special effects, and more — then proposes them for your review. This article explains how detection works, where you trigger it, and how the review-and-apply flow keeps you in control. For the click-by-click steps, see AI element detection.
How element detection works
The AI analyzes each scene’s text and matches what it finds against the breakdown element types. It uses context to tell, for example, a car that appears on screen from one only mentioned in dialogue. Detection is the upstream feed for your breakdown, so its results flow on to the schedule and reports.
Detection runs as a background job. While it runs, the AI job status bar appears below the top bar with a progress percentage and a cancel X; you can keep working elsewhere. Detection uses credits at 1 credit = 1 scene processed, charged only when the run succeeds.
Where to trigger detection
You start element detection from the AI tools menu, which appears in two places:
- The breakdown overview, acting on the scenes you select with the row checkboxes.
- A single scene’s scene detail view, acting on that scene.
In both menus the action is Element Breakdown. In the breakdown overview the label shows the count of selected scenes, for example Element Breakdown (3). Running element detection requires the Owner or Admin role and at least one selected scene.
[SCREENSHOT: The AI tools menu in the breakdown overview with Element Breakdown highlighted and three scenes selected]
For the full procedure, see AI element detection.
How the review flow works
Detection never writes to your breakdown directly. It produces a proposal that you review and apply. When the job finishes, SceneItAll notifies you with a toast, a notification in the bell menu, and a Review banner reading “AI results are ready for your review.” Each scene that has pending changes also shows a small accent pill with the change count in the breakdown.
Select Review (on the banner, the per-scene pill, or the notification) to open the Review AI Changes dialog. The dialog groups changes by scene and splits each scene’s changes into two kinds:
- New — elements the AI found that are not yet in the scene. These are checked by default; uncheck any you do not want.
- Flagged for Removal — elements already on the scene that the AI did not find this run. These are unchecked by default; check the ones you want removed.
Use Select all new or Deselect all new to toggle the discovered items in bulk. The dialog header summarizes the counts, for example “12 new · 3 flagged for removal.”
To apply your selection, use the button at the bottom right. Its label reflects what you chose, for example Apply: Add 12, Remove 3. To discard the whole proposal, select Reject All. If the AI found nothing to add or remove, the dialog shows a single Mark as Reviewed button instead — selecting it clears the pending-review banner and pill.
[SCREENSHOT: The Review AI Changes dialog showing a scene with New and Flagged for Removal sections and an Apply button reading “Apply: Add 12, Remove 3”]
Once you apply, accepted elements are created (or matched to existing ones) and tagged to their scenes; everything you left unchecked is discarded.
Reference
Proposal states
A detection proposal moves through these states as you review it:
| State | Meaning |
|---|---|
pending_review | The proposal is waiting for your review. |
accepted | You applied every change. |
partially_accepted | You applied some changes and left others. |
rejected | You discarded the proposal. |
expired | The proposal aged out before review. |
What detection reads
- The AI reads each scene’s screenplay text.
- It assigns each detected element to one of the breakdown element types. See Element types.
- It matches detected items against your existing elements before creating new ones, to avoid duplicates.
Tips
- Run detection on a few scenes first to see how it handles your script’s writing style, then process the rest in a batch.
- Clear, descriptive action lines produce better results than vague text. “He picks up the revolver” tells the AI more than “he picks it up.”
- After applying a proposal, open the tagged scenes in scene detail to catch anything the AI got wrong. You can always add, remove, or merge elements by hand afterward.