Both of these are about taking something out of a clip and carrying on with it: one takes the picture, the other takes the content. In DOTV, a frame grab turns a clip's first, last or current frame into an image node, so the next shot can be generated from that frame; video analysis has a model watch the whole clip and work its shot script back out of it — the quickest way to study, or recreate, someone else's AI video.
Move the mouse onto the video and a control bar floats up along the bottom edge. The button at the far right is the frame grab; the toolbar's Extract dropdown ends with the same entry, Grab frame, and the same three ways:
Click that camera to grab the current frame; hovering opens the three entries:
The frame lands as an Image node to the right of the original, and from then on it is an ordinary image: you can edit it, use it as a reference, and wire more onto it (see Chapter 3).
The common move is to take the last frame as the first frame of the next clip, so the two cut together continuously.
Analyze, inside Extract on the toolbar, is the other half: it sends the whole clip to a model and has it write that clip's shot script — how long each shot runs, what it shows, how the camera moves.
Click it and a draft script node lands to the right, already selected, with a read-only input bar floating up below: the prompt comes from the system (the line in the box, "Generate a script from video", is only a label), and what you pick is the model — Doubao Seed 2.0 Lite Omni (with Audio) by default, which can hear the dialogue and the score, or Doubao Seed 2.1 Pro Vision, stronger on the picture but unable to hear anything. Nothing runs until you hit send: this step is billed by the frames it samples, a real cost, so it is not started for you (and because the sampling follows the clip's length, this one shows no credit estimate). The script that comes out can be laid straight out as a full set of storyboard images, a full set of videos, a full set of voiceovers — all of that is in Chapter 6.
Film study, right beside it, is the same kind of thing with a different output: instead of a script it breaks the clip into four sets of references — Shots, Motion, Music and Script — and lays them out to the right. Clicking it drops a draft film-study node; tick which dimensions you want on the card, then press Start study — that section is 4.8.
Both are the least work when you are breaking down or copying someone else's film: get the structure out first, then change it to suit.
The same Extract dropdown holds two more: Depth capture and Cut out subject. Both run a model over the clip frame by frame on your machine — no credits, and the original is untouched. Pick a resolution (480P / 720P) and a video node lands on the right to hold the output: