Repository navigation
Importing video
Two different things, and the dialog offers both because they are wanted for different reasons.
Where: Media → import a video file, or paste a YouTube link (the link route needs the local server running, since the browser cannot fetch YouTube itself).
The video is chopped into frame cuts you draw over, then the video is thrown away. This is the one to pick when you are tracing.
Frames a second. How many cuts you get per second of video. Four is plenty for tracing; the higher you go, the more cuts and the more memory — see Working with a lot of cuts.
Scale. Frames are shrunk on the way in. Half size is usually enough to trace from, and it is a quarter of the memory.
Quality. Compressed WebP is the default and is small. High-quality WebP is near-lossless at the original resolution. Lossless PNG keeps every pixel and is large — only pick it if you are going to use the frames as-is rather than draw over them.
Only a range. Start and end in mm:ss, so a three-minute video does not become nine hundred
cuts when you wanted eight seconds of it.
Merge duplicates. Video has a lot of frames that are the same as the one before — a held shot, a title card — and each one would otherwise be its own cut.
- Exact merges only pixel-identical frames. It checks a small signature first and then compares every byte, so it never merges two frames that differ at all. On this setting the only mistake it can make is not merging two frames a person would call the same, because compression noise made them differ by a few bytes.
- Nearly the same and Loose merge frames whose signature is within a threshold. More merging, fewer cuts, and the risk goes the other way.
A merged run becomes one cut held for that long, so the timing is preserved.
Split into parts. A long video comes in already grouped into parts, one every thirty seconds or so; the dialog suggests a count from the video's length. Parts are what the split-and-rejoin workflow works on.
Canvas size. Match the video, use the standard landscape or the vertical shorts size, or keep whatever the canvas is now. A frame that is a different shape is fitted, not stretched.
With the audio. The video's own sound comes in as the music track, aligned to the first imported frame and clipped to the range you took.
The whole video is laid under your layers, playing along with the film, and never becomes cuts. This is for reference or for a shot where the video is part of the picture.
It has an opacity, in the same settings as the camera, and it goes through the camera transform — so a camera move moves the video with the drawing, which is what you want if it is part of the shot.
On a reference track, the app can find the cuts in the video — the moments where one shot becomes another — and put markers on the timeline so you can jump between them. The threshold is how big a change counts as a cut; lower finds more.
It runs in the background and can be stopped. If you open another project while it is running the result is thrown away, because it describes a video that is no longer loaded.
Extracting frames takes a while for a long video. Send to background closes the dialog and carries on in a corner chip — you can keep drawing. Opening a project, on the other hand, cannot be done in the background, because it replaces the document you would be drawing on.
Recently imported videos are remembered for the session, so re-importing the same one with different settings does not download it again.