Get the frames worth keeping from any video

How to extract frames from a video

A video frame is a single still picture out of the thousands that make up a video. Paste a link or upload a file, and this tool pulls those stills out for you — as images you can download, use as thumbnails, put in a storyboard, or search the web with.

The difference is in which frames you get back.

Why "every N seconds" is the wrong question

Almost every frame extractor works the same way: you pick an interval, it slices the video at that interval, you get a folder of pictures. Slice a ten-minute video every two seconds and you have three hundred images. Most of them are near-identical to their neighbours. A good number are smeared by motion blur. Several are solid black, because they landed in the middle of a fade.

Nobody wants three hundred frames. They want the twenty that are worth having. Choosing those is the actual work, and it is the part every other tool leaves to you.

What this tool does instead

Before a single frame is delivered, the video is sampled across its whole length and every sample is measured:

• Sharpness — how much fine detail is present. This is what rejects motion-blurred frames.
• Exposure — whether the frame is crushed to black, blown out to white, or sitting comfortably in between.
• Contrast and colour — how much range the picture actually uses.
• Detail — how much visual information the frame carries, which is what separates a busy scene from a plain wall.

Each sample is also fingerprinted, so frames that show the same picture can be recognised as the same picture and merged. When the fingerprint changes sharply, that is a cut — which means the shot boundaries come out of the same work, for free.

The five modes

• Smart storyboard. One frame per shot, and the best frame in that shot. This is the default, and for most videos it is the answer: a real storyboard, one picture per scene, no repeats.
• Best thumbnails. The highest-quality frames in the video, spread across its length so you do not get five views of the same second. This is the mode for choosing a cover image.
• Reverse-search ready. The frames most likely to work as a probe for finding this video elsewhere online.
• Fixed interval. A frame every few seconds, the classic behaviour — kept because it is genuinely the right answer when you need a set cadence. Duplicate merging is optional here.
• Single moment. One frame at one exact timestamp, when you already know the moment you want.

Quality and reverse-search are not the same score

Every frame comes back with two numbers, and they measure different things.

Quality answers "is this a good picture?" — sharp, well exposed, good contrast and colour. It is the score to sort by when you are picking a thumbnail.

Reverse-search answers a different question: "would this frame find the video if I searched the web with it?" That depends on distinctiveness, not prettiness. A perfectly sharp, perfectly exposed frame of a blank wall is a lovely picture and a useless search probe, because there is nothing in it for a search engine to match. Fades, near-black frames and title cards score badly for the same reason.

The two scores often disagree, and when they do, the reverse-search number is the one to trust for tracing a video back to its source.

Why only we can offer that

Scoring a frame on how well it would reverse-search is only meaningful if there is a reverse video search to send it to. This tool sits on a site that runs one, so every frame carries a one-click search action — take the most distinctive frame from a clip and find where else it has been posted, without leaving the page. A standalone frame grabber has nothing to hand the frame to.

What you can do with the frames

• Download them all as a ZIP. Built in your own browser, so nothing is stored on our servers at any point.
• Build a contact sheet. Every frame on a single image with its timestamp burned in — a storyboard you can send to someone.
• Take one frame. Every frame has its own download, named with the exact timestamp it came from.
• Search the web with a frame. Send the most distinctive frame into a reverse video search and find where else the clip appears.

Choosing a size and a file type

Frames are delivered at the width you choose, up to the video's own resolution — asking for 1920px from a 720p source returns 720p rather than a soft blow-up pretending to be HD.

• WebP is the default: the smallest files at a given quality, and supported everywhere that matters.
• JPEG for maximum compatibility with older software.
• PNG when you need lossless frames for editing. These are much larger files.

Who uses this

• Creators pulling a thumbnail out of footage they already shot, instead of staging a separate photo.
• Editors and producers building a storyboard or a shot list from a rough cut.
• Journalists and researchers taking a clear still out of a clip to verify it, publish it, or trace it back to its source.
• Anyone tracing a video who needs the one frame most likely to turn up matches in a reverse image or video search.
• Teachers and course builders illustrating written material with stills from a recording.

Privacy

Your video is read, the frames are cut, and the file is deleted — nothing about it is kept. The frames themselves are sent straight to your browser and are never written to disk on our side. The ZIP and the contact sheet are both assembled on your own machine, so those images never travel anywhere at all.

Frames, keyframes and thumbnails

A frame is any single still in the video. A keyframe is a technical term from video compression — a frame stored in full rather than as a difference from the one before it. Encoders tend to place keyframes at cuts, which is useful, but a keyframe is not automatically a good picture. A thumbnail is whichever frame you decide represents the video. This tool helps with the last one: it finds the frames that would make good thumbnails, rather than simply listing what the encoder happened to store.