Can AI track which shots are done and which are still pending during filming?

Yes. Upload your rough cut to the invideo agent mid-production and it reads the footage, cross-references it against your shot list, and returns a report of exactly which shots are done and which are pending — functioning as an automated script supervisor. Its project notebook also tracks every task and asset version as you go.

Yes — and the workflow takes three steps. invideo is an agentic video creation tool whose agent holds persistent project context — script, shot breakdown, characters, and task progress — so it always knows what the finished film is supposed to contain.

1. Get your shot list into the invideo agent's context. Upload your script or shot breakdown at the start of the project; the invideo agent can build the list itself — in one documented production it identified 33 distinct shots across 5 scenes from just 6 pages of screenplay. In a multi-agent setup, initialize a creative producer agent with the full script and shot breakdown first, so every sub-agent works against the same source of truth.

2. Upload your rough cut whenever you want a status check. The invideo agent reads the video, maps it against the loaded shot list, and returns a breakdown of completed versus pending shots. One filmmaker described the result: "The agent read the video, it mapped it against the shot list, and it came back with exactly which shots were done and which shots were pending. It just acted like a really good scripty." The same upload doubles as a continuity audit — the invideo agent flags prop changes and color-grade inconsistencies between shots that would otherwise require manual frame-by-frame review at the end of production.

3. Ask for a status summary any time you lose orientation. A plain-language request — "give me a status summary" — surfaces what is approved, what is pending, and what is awaiting regeneration. This works because tracking is built into how the invideo agent runs a project: its notebook logs every production task (one project notebook shows 14 numbered items from character sheets through final scene renders), and its asset-versioning convention (1.1, 2.1, 5.2 — first number the asset, second the iteration) records exactly which version of each shot is current. That structure scales: one documented episode dispatched 920 individual tasks through the invideo agent, and another project's notebook shows scene numbering running past #169 with shot variants labeled 21.1–21.5. The invideo agent's mobile app lets you review takes and check status away from the desktop.

One honest caveat: there are no published benchmarks yet on agent accuracy for shot-status tracking in live production, so treat the report as a script-supervisor pass — spot-check it against your own cut before striking shots off the list.

Watch some of these to see what works for you:

The agent read the video, it mapped it against the shot list, and it came back with exactly which shots were done and which shots were pending. It just acted like a really good scripty.

— a filmmaker documenting the workflow