What is reverse video search?And how it actually works
Reverse video search means using a video itself as the search query. Instead of typing words that describe it, you give the search the footage, and it looks for web pages that show the same images.
It is how you find where a clip came from, where else it has been posted and whether it is older than it claims. In practice, it works by turning the video into a set of still frames and searching each one visually, because search engines index images far better than they index video.
Searching by content, not by words
A normal search matches the words you type against the words on web pages. That fails with video, because a reposted clip often has a different caption, a different language or no text at all. The footage is the only thing that stays the same.
Reverse search flips the process. You start from the content, and the search looks for pages that display visually similar images. It does not need to know what the video is about, who is in it or what it is called.
You may know reverse image search already, with tools such as Google Lens or TinEye. Reverse video search builds on the same idea, with extra steps to deal with the fact that a video is thousands of images in a row.
How it works, step by step
Most reverse video search, whether you do it by hand or with a tool, follows the same three steps.
- 1.Extract keyframes. The video is sampled into still images. Good samples are distinctive moments: a clear face, a recognisable place, a sign, an unusual action. Blurry transitions and black frames are useless. You can do this yourself; see how to extract frames from a video.
- 2.Search each frame visually. Each still is sent to a visual search engine, which compares it against the images it has indexed from the web and returns pages that show the same or a very similar picture.
- 3.Group the pages that match several frames. One frame on its own can match lots of unrelated pages, such as stock photos or similar scenes. A page that matches three or four different moments of the same clip is very likely to show that clip. Grouping results across frames is what separates real copies from coincidences.
The search finds pages that show the images. It does not watch videos on those pages, read their upload dates or know which copy came first. You compare that yourself on the pages it finds.
Why it is harder than image search
- —Search engines index images, not video. What they mostly see of a video page is its thumbnail and a few preview images. If none of your frames looks like those, the page may not come up.
- —One frame is a lottery. A 30 second clip contains hundreds of frames, and the one you pick may not be the one that was indexed. Searching several moments raises your chances a lot.
- —Reposts change the picture. Copies get cropped, zoomed, mirrored, filtered, covered with text or squeezed into a split screen, and every change makes the match harder.
- —Most video lives behind apps. Private accounts, stories, chats and closed groups are invisible to search engines, and so to any reverse search.
What it can and can't find
Reverse video search only finds public pages that a search engine has indexed. Within that limit, here is what to expect.
- —Usually found: reuploads of the same footage on YouTube, TikTok, Instagram, X, Facebook, Reddit, news sites and blogs, as long as those pages are public and indexed.
- —Sometimes found: copies that were cropped, mirrored, lightly filtered or have text added. Enough of the image often survives for at least some frames to match.
- —Rarely found: heavily edited versions, such as clips used in a small corner of a reaction video, strongly recoloured footage or remakes filmed again.
- —Never found: private accounts, deleted posts, content in chat apps and anything a search engine has not indexed. A deleted TikTok may still survive as a copy elsewhere; see how to find a deleted TikTok.
No results does not mean no copies exist. It means none were found on pages a search engine has indexed.
What people use it for
- —Finding the original. Collect every copy, then compare dates, quality and accounts: how to find who posted a video first.
- —Checking old footage passed off as new. An earlier copy proves a clip is older than claimed: how to check if a video is old.
- —Checking a viral or AI video. Finding the source is often the quickest way to tell if a clip is real or generated: is this video AI and is a viral video real.
- —Finding your stolen videos. Creators use it to find reposts of their work: how to protect your videos from theft.
- —Identifying a movie or a meme. A clip can lead you to the film or the first post it came from: find a movie from a clip and find who made a meme.
Tools you can use
- —Google Lens, one frame at a time. Free and good at matching images. You take a screenshot, search it, and repeat for other moments. Lens has no option to upload a video, so grouping the results is up to you. See how to reverse search a video on Google.
- —InVID/WeVerify plus manual search. The free InVID/WeVerify plugin extracts keyframes from a video link and gives you quick links to search each one on several engines. It is built for journalists and fact checkers, and it still leaves you to compare the results.
- —Dedicated tools such as Retrace. Retrace does the three steps for you: upload a video and it samples up to 24 distinctive frames across the whole clip, searches them visually and groups the pages where the same footage appears. The free scan tells you how many matching pages exist and on which platforms; unlocking the search for $0.90 lists every link, the moment that matched and how many frames agree. See pricing for details.
Every option depends on the same indexed web, so they share the same blind spots. If one tool finds nothing, trying another frame or another engine can still help.