Extract frames from video

Individual frames from video are useful in more contexts than thumbnail generation alone. Training a computer vision model requires still images, not video. Creating a sprite sheet for a game or animation requires a set of sequential frames at regular intervals. Reviewing footage frame by frame for motion analysis, sports technique, or quality control means exporting images rather than scrubbing a timeline. Building a stop-motion storyboard from live footage starts with extracted frames.

Frame extraction is also a practical fallback when a video tool you are using does not support a particular format: extract the frames as JPEGs, process them as images in whatever tool you have, and reassemble if needed.

Extract Frames does not use FFmpeg at all. The extraction happens entirely through the browser's HTML video API and the canvas element. A hidden video element is loaded with the file and seeked to each target timestamp using video.currentTime = t. The seeked event fires when the browser has decoded and painted that frame. At that point, the frame is captured by drawing the video element onto a canvas with ctx.drawImage(video, 0, 0, canvas.width, canvas.height) and then exported as a JPEG using canvas.toBlob at quality 0.92. The canvas dimensions match video.videoWidth and video.videoHeight exactly, preserving the native resolution.

Two extraction modes are available: interval mode steps through the video at a fixed time interval you specify, and count mode distributes a fixed number of frames evenly across the full duration. Both modes cap at 30 frames per extraction.

The 30-frame cap exists because each frame must be individually decoded and painted by the browser's video renderer, which is slower than batch FFmpeg frame extraction. For large extractions (entire video at 1 fps for a long clip), use FFmpeg directly if you have it, or split the video first and extract from shorter segments.

JPEG quality is fixed at 0.92, which is high enough for most uses. If you need lossless frames (for compositing or further processing), use the Grab Thumbnail tool instead, which exports PNG at lossless quality.

Extraction accuracy depends on the browser's video seek precision. Most browsers seek to the nearest keyframe and then step forward to the exact timestamp, but very short intervals (under 0.1 seconds) may produce duplicate frames if the browser cannot seek that precisely.

Desktop tools like FFmpeg extract frames faster and have no frame count limit, but require a command like ffmpeg -i input.mp4 -vf fps=1 frames/%04d.jpg, which has to be looked up and adapted for each use case. Premiere and Resolve can export frame sequences from a timeline, but require a project. This tool extracts frames by seeking the video element directly, which means it works on any format the browser can decode natively (including formats FFmpeg WASM does not support) and downloads each frame as a named JPEG.