Extract video frames

One every intervalSec seconds, which defaults to 1, stopping at maxFrames, which defaults to 60. So the default settings give you at most a minute of one frame per second. Raise maxFrames if you need more, but remember every frame is a separate PNG held in memory until you download it.

Your video name with a three digit counter and a .png ending, for example clip-001.png, clip-002.png and so on, in time order, so they sort correctly in any file manager. PNG is the only output format here. For a single frame in PNG or JPEG, use Video thumbnail instead.

So a long video cannot silently produce thousands of files. Sampling an hour long recording once a second would mean 3600 PNGs built in your browser memory at the same time. The cap stops early instead, and you can raise it deliberately when you know what you are asking for.

Because no sample time fell inside the file. The tool builds the list of timestamps by stepping from zero to the video duration by intervalSec, and if that list is empty it refuses with the duration and the interval both named. Lower intervalSec so that at least one sample lands inside the clip.

Because decoding frames needs WebCodecs, which Node does not implement. The CLI and MCP surfaces refuse it with "video-extract-frames needs a browser (OffscreenCanvas). Use it at /video/video-extract-frames instead." The web page runs the same code in a Web Worker. Everything still stays on your device, since that worker runs inside the page you have open.

No. The file is decoded in your own tab and the PNGs are built there too. Nothing reaches a server. Both the source video and every extracted frame are held in memory at once, so a long video with a high maxFrames is the case most likely to run a tab out of room.