AI Autofill Pro
AI Autofill looks at one frame from each shot and writes a value into the shot metadata fields you switched it on for. It fills the fields you enabled — nothing else — and every field ships with its Autofill switch off.
AI Autofill is available on the Pro plan. See Compare plans.
What can AI Autofill fill?
Shot metadata fields you created and enabled, of any field type. What comes back depends on the type:
| Field type | What comes back | Held to your options? |
|---|---|---|
| Dropdown | One of the options you defined | Yes |
| Rating | A whole number from 1 to 5, or 0 for unrated | Yes |
| Checkbox | Ticked or unticked | Yes |
| Number | Any finite number | Yes |
| Tag | Short tags, which can include ones you never listed | No — open on purpose |
| Text | Anything | No — a pass-through |
That split is the thing to remember. A Dropdown can't come back with an option you didn't write. A Tag or Text field can come back with anything, because a tag list that could only repeat itself would never grow — which also means a Job# in a Text field is exactly where a plausible wrong answer has nothing to stop it.
Two things it never does:
- It never creates a field. That's AI field discovery, a different feature that proposes fields and fills no values.
- It never touches a video-level field. Autofill is a shot-metadata feature, so Preferences → Video Metadata carries no Autofill control at all. See Video metadata.
How do I turn it on for a field?
- Open Preferences (Cmd+, on Mac, Ctrl+, on Windows).
- Go to Notes & Metadata → Shot Metadata.
- Find the field you want filled and turn on its Autofill switch.
- Give the field a one-line description while you're there. It's the cheapest steering available — Shot Size with the description "framing of the main subject" produces different answers from Shot Size with nothing.
Start with one or two fields. The Autofill action doesn't appear anywhere in the app until at least one field has its switch on, so if you can't find it, this step hasn't happened yet.
What are the two global rules?
They live in Preferences → Notes & Metadata → AI Autofill, and they apply to every run.
- Autofill on import — run Autofill on the enabled fields as soon as a video finishes importing. Off by default. Leave it off until you trust the schema: a long video is hundreds of model calls from one drag-and-drop.
- Only fill empty fields — Autofill never overwrites a value a person wrote. On by default, and it's the main guard you have.
How do I run it?
From the inspector, in the editor or the library.
- Select the shots you want filled.
- In the inspector, choose Autofill with AI….
- The Autofill fields with AI dialog opens on your enabled fields, all ticked. Untick any you don't want in this run.
- Pick the scope — This shot, N selected, or All N shots.
- Pick the write mode — Only fill empty fields or Overwrite existing values. Overwriting shows you a count of what it would replace before you commit.
- Add run instructions if you want to steer this run only. They aren't saved.
- Read the count line — it names the number of fields and shots — then click Autofill N shots.
The dialog launches the run and closes. Progress shows as per-field spinners on the shot card and in the inspector, not as one bar. The run is a queued job: you can navigate away, and it survives restarting the app.
What leaves my computer?
One frame per shot, and the context around it.
StoryFolder sends each shot's frame to StoryFolder's server, which passes it to a hosted vision model along with your field names, their types and options, your field descriptions, this run's instructions, and the video's title, description and video-level field values.
Your video file is never uploaded. But this is a cloud feature, and it needs an internet connection every time. That's the opposite of transcription, which runs entirely on your machine. Don't read "local" as an app-wide promise — see Local vs cloud and Privacy & sharing.
What are the limits?
- There is no undo. An Autofill run doesn't create a restore point, so version history can't roll one back. Your guards are Only fill empty fields and running small before you run wide.
- A valid value isn't necessarily a right value. A Dropdown answer that's one of your options is accepted whether or not it's the correct one of your options. That's the failure you'll actually meet.
- A rejected value is silent. If the model answers outside a field's domain the value is dropped and nothing is written — so an empty cell doesn't tell you whether the model declined or was overruled.
- Nothing marks a value as AI-written. In the cell, an Autofill value looks exactly like one you typed.
- It can't know what isn't in the frame. Client, job number, release status and who booked the day live in your head or in an email. Put those in video metadata or paste them across shots.
How do I review a big board?
Filter, don't scroll. For each closed field, filter to one option at a time and look at that queue — Shot Size = Wide, then the next option. Open Text and Tag fields have no queue to filter into, so take a fixed sample from the beginning, middle and end, plus the shots most likely to defeat a single frame: dark shots, transitions and graphics. Fix what's wrong in place, then widen the run.