Video Frame Capture gets still images out of a video: a cover image, illustrations for a written tutorial, or a run of frames taken from a stretch of footage. It is the “Frame capture” tab of the audio/video workbench and holds two methods that do not interfere with each other.

“Capture current frame” does not use ffmpeg at all: pause the preview at the frame you want and click once, and the browser draws the player's current picture onto a canvas as a PNG at the video's original resolution — no transcoding, done instantly, with the time point in the file name such as <original-name>_00-12.png. “Batch capture by interval” goes through ffmpeg's fps=1/interval filter to take one frame every N seconds and outputs PNG (lossless) or JPG (-q:v 2, high quality); when all the images are ready, Download packs them into a zip whose files are numbered <original-name>_frame0001.png and onwards.

1. 01Add a video and select it— switch to the “Frame capture” tab.
2. 02Single still— drag the progress bar in the preview to the target picture and pause, then click “Capture current frame” and the PNG downloads directly. For formats the browser cannot preview (some MKV and AVI, for instance) this button is disabled.
3. 03Batch capture— fill in “Interval (seconds)” (minimum 0.2, default 5) and choose PNG or JPG; the panel estimates “about N images will be output” and warns about memory pressure above 200.
4. 04Run the batch and download— click “Start batch capture”, then Download for the zip when the job finishes.

Interval floor of 0.2 seconds

— every single frame cannot be exported (that is what the GIF Toolbox frame splitter does for animations) and there is no smart extraction based on scene changes

Count and memory

— all frames are held in memory before being packed, with a warning above 200; a 1080p PNG runs to several MB each, so widen the interval or switch to JPG

The single still needs a preview

— codecs the browser cannot decode (HEVC, AV1, some AVI / MKV) cannot use “Capture current frame” and only batch capture works; batch capture does not support AV1 either

Time precision

— the first batch frame comes from around 0 seconds and the rest follow every N seconds; a time list such as “second 3, second 17” cannot be given

What it does not do

— no thumbnail grid stitched from several frames, no timestamp watermark, no GIF export (use Video to GIF)

The single still is produced by the browser canvas directly and batch capture runs on the local ffmpeg.wasm, and the video is not uploaded in either case; the first batch capture downloads about 31 MB of engine from a CDN (the site's asset domain or jsDelivr) and that request contains no video content.