To reverse image search a video, pause on a sharp, distinctive frame, save it as a still image (screenshot, VLC snapshot or ffmpeg), crop away player controls and captions, and search that still in Google Lens and TinEye. Search two or three different frames, then use the matches, titles and upload dates to find the earliest upload of the full video. A matching frame shows the footage was reused; it does not prove the original date, place or context by itself.
The quick method
- Save the video's URL, the account name, the caption and the timestamp you plan to use.
- Pick 3 to 5 frames: an establishing wide shot, a frame with readable text or a landmark, and a frame with the clearest view of the main subject.
- Capture each frame at full size, with the video in full-screen or at its highest quality.
- Crop out player buttons, progress bars, app overlays, subtitles and watermarks added by reposting accounts. Keep a separate uncropped copy.
- Search each frame in Google Lens and TinEye. Add Bing or Yandex if both come up empty.
- Open video results and compare the full sequence, not just the one frame.
- Find the earliest credible upload and read its title, description and date.
Pick frames that search well
Search engines match stills, so the frame you choose matters more than the engine.
| Good frames | Poor frames |
|---|---|
| The first wide shot, before any zoom | Transition frames, fades and cross-dissolves |
| Signs, shopfronts, number plates, posters, landmarks | Motion blur, fast pans, shaky handheld moments |
| A distinctive object or skyline | Dark or heavily compressed night scenes |
| The video's own thumbnail or cover image | Frames covered by captions, stickers or emoji |
| A face looking at the camera (for face search) | Faces in profile, far away or partly hidden |
The first second and the thumbnail are worth extra attention. Reposters often keep the original's opening frame and thumbnail, and these are the frames search engines are most likely to have indexed.
Grab a frame from YouTube
- Desktop: pause the video, switch to the highest quality under Settings (gear) > Quality, and go full-screen. Use the . (full stop) and , (comma) keys to step forward and back one frame at a time while paused. Move the mouse away so the controls fade, then take a screenshot.
- Thumbnail trick: every public YouTube video has a thumbnail at
https://img.youtube.com/vi/VIDEO_ID/maxresdefault.jpg, whereVIDEO_IDis the code afterv=in the URL (or afteryoutu.be/). If the maxres version is missing, tryhqdefault.jpg. Searching the thumbnail often finds other uploads of the same clip. - Chrome: right-click the page and choose Search with Google Lens, then drag a box around the paused frame. This searches what is on screen without saving a file.
- Mobile: pause, wait for the controls to disappear, and take a screenshot. On many Android phones, Circle to Search works directly over a paused video.
Grab a frame from TikTok, Instagram Reels, X and Facebook
These apps do not offer frame stepping, so work with pauses and screenshots.
- Pause, then wait a moment for the on-screen buttons to fade where the app allows it. On desktop web versions, the controls usually hide when the mouse leaves the video.
- Use the desktop site when possible: the video plays at a larger size, which gives you a sharper screenshot than a phone screen.
- Crop out the interface: like and share buttons, usernames, captions and the reposter's watermark. A frame full of app chrome mostly matches other screenshots of the same app.
- Note the on-screen watermark before you crop it. TikTok exports carry the uploader's handle, which is a lead to the account that posted it, though not proof they filmed it.
The screenshot search guide covers cropping and cleaning screenshots in more detail.
Grab a frame from a video file
On a phone
On iPhone, open the video in Photos, pause, scrub to the frame and take a screenshot (side button and volume up together). On Android, pause in your gallery or video app and screenshot with power and volume down; some gallery apps also have a capture or save frame option in the editor. Then crop in the Photos or gallery editor.
VLC (Windows, Mac, Linux)
VLC saves a frame at the video's full resolution, which beats a screenshot of a small window.
- Open the file and pause on the frame. Use E to step forward one frame at a time.
- Choose Video > Take Snapshot (Windows and Linux) or Video > Snapshot (Mac).
- Find the image in the snapshot folder, which you can set under the video preferences. By default it goes to your Pictures folder on Windows and Linux and the Desktop on a Mac.
ffmpeg (any platform, best for many frames)
ffmpeg is free and extracts lossless frames from almost any format.
One frame at a specific time:
ffmpeg -ss 00:01:23 -i clip.mp4 -frames:v 1 frame.pngOne frame per second, to pick the best ones by eye:
ffmpeg -i clip.mp4 -vf fps=1 frame_%04d.pngOnly frames where the scene changes (usually the most useful set):
ffmpeg -i clip.mp4 -vf "select='gt(scene,0.3)'" -fps_mode vfr scene_%03d.pngOlder ffmpeg versions use -vsync vfr instead of -fps_mode vfr. Lower 0.3 to catch more cuts, raise it to catch fewer. If you suspect a repost was mirrored to dodge detection, flip a frame back with ffmpeg -i frame.png -vf hflip flipped.png and search both.
Screenshot shortcuts
On Windows, Win + Shift + S opens the snipping tool for a region. On a Mac, Shift + Command + 4 lets you drag a box; see the Mac reverse image search guide for searching it straight away.
Search the frame
Search every frame in at least two engines and keep notes per frame: timestamp, engine, and the best candidate URLs.
| Engine | Why use it for video frames |
|---|---|
| Google Lens | Widest coverage, often returns YouTube and news video pages directly; use Exact matches |
| TinEye | Finds exact and edited copies of the still; sort by Oldest |
| Bing Visual Search | A second large index; sometimes surfaces different video pages |
| Yandex Images | Can find copies on sites and regions the others miss |
If nothing comes back, try a tighter crop on the most distinctive object, a different frame, or the thumbnail. Heavy filters, zooms, text overlays and mirroring all reduce matches. The not-working troubleshooting guide lists more fixes.
Fact-checkers also use the free InVID-WeVerify browser plugin, which can split a video link into keyframes and send them to several engines at once.
Find the original video
A frame match gets you to other copies. To get to the first upload:
- Collect every video result from your frame searches, plus pages that embed the clip.
- Compare the footage itself. The original is usually longer, higher resolution, has no reposter watermark, and may include a few seconds before and after the version you saw. A shorter, cropped, captioned copy is almost always a repost.
- Check upload dates on the platform. YouTube shows the upload date under the title in the description area. On TikTok and Instagram the date appears with the caption, sometimes as relative time ("3d ago"); hover or open the post page for the full date where shown.
- Search text clues. Put distinctive words from signs, the original caption or a translated description into the platform's own search and into a web search. On YouTube, use Filters > Upload date and sort by upload date to see older copies first.
- Look at the uploader. An account with other footage from the same place and time, or the raw longer version, is more likely to be the filmer. A compilation or news-aggregator account is not.
- Check archives. If the earliest copy was deleted, the Wayback Machine may have the post page.
Write your result as "the earliest upload I found was posted by X on [date]". Earlier copies may exist in private chats, deleted accounts or places search engines do not index.
Check the claim attached to the clip
The commonest problem with viral video is old or unrelated footage attached to a new event. Once you have the earliest upload, compare its title, description and comments with the claim you are checking: date, place and event. Look for clues in the footage too: language on signs, weather and season, vehicles and uniforms. Two storms over the same city can look nearly identical, so rely on exact matches in the sequence, not general resemblance. The full method for testing captions is in how to verify a photo before you share or publish it, and tracing a still back to its creator is covered in finding the original source of an image.
Identifying a person in a video, responsibly
Sometimes the question is not where the footage came from but who is in it, for example a dating match who sent a video, or a clip of you posted by an account you do not know. That needs a face search rather than a copy search.
- Pick the frame carefully: face toward the camera, eyes visible, no motion blur, good light. Crop to the head and shoulders. Try two or three frames, since a single frame can be unflattering or soft.
- Expect lower accuracy than a photo. Video compression, low resolution and filters make face matching harder. Treat every match as a lead and open the source page to check it.
- Watch for beauty filters and AI-generated video. Both change facial features. The AI image and deepfake checks help when the footage itself is suspect.
Facehunch runs this kind of face search for $4.99 per search, with each match shown alongside its source link and a similarity score. Use it to check someone you are already talking to or to find videos and images of yourself. Do not use it to identify strangers filmed in public, work out where someone lives, or contact someone who does not want contact.
Written by Facehunch with AI assistance and checked against the cited sources. Our editorial policy.