A video node holds the core output of AI video creation — a clip: generated by a model, or existing footage you uploaded yourself. In DOTV it uses the same card as an image node, except the picture moves — sweep the mouse across it and it starts playing. This section covers what's on the card, what you can do with it selected, and why "Edit frame" edits exactly one frame.
An uploaded clip shows a blurred preview while it transcodes and transfers, with the line above it reading "Transcoding… 42%" first (only clips that are not already H.264 need it, and the percentage runs to the end) and then "Uploading…". That is not a failed generation. A real failure puts a plain-language line in the frame, with Retry next to it.
Once the Agent has watched this clip and transcribed it, a Dialogue badge hangs after the name. Open it and you get a read-only script line by line, in time order, each line prefixed with the minute and second it falls on — when it does not match the picture, one glance tells you whether the dialogue was misheard or the shot is off.
That script is kept on the clip itself (transcribing again gives you a new one), and it is only fetched when you open it, so it costs almost nothing the rest of the time.
None of these change the original clip: trimming, upscaling, subtitle removal, subject edits, reshoots, extensions, extractions and audio separation all produce a new node to the right, and the original stays exactly as it was, prompt and all.
Select a video node that already has a clip and this bar floats above the picture:
| Button | What it does | More |
|---|---|---|
| Trim | Cut one stretch out of it | 4.3 |
| Reshoot clips ▾ | Two items in the dropdown: Reshoot clips marks the stretches to change and regenerates the whole clip; Smart continue plays on from the stretch you marked | 4.3 |
| Edit ▾ | The things you do to this clip that still give you back a clip: Upscale, Crop, Remove subtitles (smart or boxed), Erase / Change / Replace subject. The Edit frame group at the end is the odd one out — it captures the frame on screen, so what you get is an image node | 4.4 / 4.5 |
| Extract ▾ | Take something reusable out of this clip, onto a new node to the right: Analyze, Film study, Depth capture, Cut out subject (the last two each with a resolution), Audio separation (three ways to split). The Grab frame group at the very end (current / first / last frame) cuts a frame into an image node | 4.6 / 4.7 |
| Publish effect | Send this clip's look, along with the settings it was made with, to the effect library | 4.2 |
| Download | Save it locally. The whole clip has to be fetched first, so tens of MB means a wait — the button shows what percentage has been fetched (with no length reported by the server it only spins); when the browser blocks the automatic download, you get a line and a Save file now button | — |
| Play fullscreen | Enters native fullscreen immediately, with the full controls; press Esc to exit | — |
The Edit frame group at the end of the Edit dropdown comes in two flavors: Inpaint (paint over an area and have it redrawn from your prompt) and Expand frame (grow the canvas and fill the new margins). Both work on the frame currently on screen: clicking one captures that frame into a new image node and opens the matching editor on it — paint the mask or drag the frame, confirm, and the edit lands. The editing itself is the same machinery as on image nodes, see 3.3.
To get an "edited video" out of it, use the edited frame as the first frame of a fresh video generation (see 4.2).