Built on advanced on-device AI and hardware acceleration to deliver professional-grade performance with uncompromising privacy.On-device AI — professional-grade, fully private.
Video Thumbnail Generator Online
Video thumbnails from any frame — pick the exact moment.
- No upload — runs on your device
- PNG, JPEG or WebP
- Up to 50 frames in one pass
Drag & drop a video
or click to browse — MP4, MOV, WebM, MKV, GIF
Platform sizes, or the video's own frame.
Up to 5 on this video.
How the frame meets the target box when the shapes differ.
Every capture tiled into a single image.
Why generate video thumbnails?
Extract still frames from any video to use as thumbnails. Compatible video in, still images out: one or many timestamps, custom dimensions, single files or a ZIP, and private processing that never uploads your footage.
A thumbnail is the only part of a video most people ever see. It decides whether a viewer clicks, it fills the card in a feed or a course index, and it is the still that represents the whole clip everywhere the clip itself cannot play. Extracting one from the footage keeps the picture honest — the cover really is a moment from the video — and costs nothing but the time it takes to find the right frame.
Benefits of video thumbnails
- Preview image: Show the key moment before playback, so a viewer knows what they are choosing.
- Platform fit: Prepare assets at the exact sizes YouTube, Shorts and link previews ask for.
- Organization: Index and browse a video library faster when every entry has a picture.
- Brand control: Pick the frame that best represents your content rather than accepting an automatic one.
- Reuse: Use captures in pages, posts, documentation and marketing assets.
Thumbnail generation explained
Capturing a thumbnail is three operations: find the moment, decode the picture that is there, and write it out at the size you need. The middle one is where the time goes, because a video is not a stack of images — most frames are stored as differences from an earlier one, so reaching any given moment means decoding forward from the nearest complete frame before it.
Generation process
- Read the container: Parse the file's structure and find the video track, without decoding a picture.
- Sort the moments: Put every requested timestamp in order, so the clip is read once instead of once per capture.
- Decode in one pass: Walk the video forward and take a picture at each requested moment as it goes by.
- Fit to the box: Crop, letterbox or stretch the frame into the size you asked for, shrinking in stages so fine detail survives.
- Write the image: Encode PNG, JPEG or WebP and hand it back as a file.
Choosing the right frame
The difference between a thumbnail that earns a click and one that does not is usually a few frames, and the frames either side of a good one are often unusable — a half-closed eye, a hand across a face, the motion blur in the middle of a turn. Choosing well is a matter of being able to move in small enough steps and to compare several candidates.
Working frame by frame
- Scrub to roughly the right moment first, then step single frames until the expression lands.
- One step is one frame at the video's own rate, so a 24 fps film and a 60 fps screen recording both move by exactly one picture.
- The preview shows the finished crop and size, not the raw frame, so nothing shifts between what you choose and what you download.
- Avoid the very first and very last second of a clip: openings fade up from black and endings fade down to it.
Comparing candidates
- Ask for several captures and space them evenly to see the shape of the whole clip at once.
- A contact sheet puts every capture in one image, which is the quickest way to review a long recording.
- Generate three or four options and compare them at thumbnail size — a picture that reads at full size often does not read at 210 pixels wide.
- High contrast and a clear subject survive being shrunk; fine text and busy backgrounds do not.
Sizes, formats and fit
Every platform states a thumbnail size, and most of them reject anything that does not match. The four presets here are those sizes; the fit mode decides what happens when the video's shape and the target shape disagree, which they usually do.
Quality considerations
- Keep the output at or below the source resolution — enlarging cannot add detail that was never recorded.
- Choose PNG for sharp text and graphics, JPEG for photographic frames, and WebP when file size matters most.
- Crop fills the target and trims the overflow from the centre; letterbox fits the whole frame and pads the remainder.
- Stretch distorts deliberately, and is only ever the right answer when a platform demands a shape and the distortion is intended.
- Shrinking happens in stages rather than in one jump, which is what keeps texture and small text from breaking up.
Common use cases
- YouTube uploads: Custom thumbnail candidates from key scenes, at the 1280×720 the platform asks for.
- Social media posts: Optimised static previews for feeds, at square and vertical sizes.
- Course and library indexing: Consistent chapter and lesson pictures across a whole catalogue.
- Product demos: Highlight the exact feature moment from a screen recording.
- Editorial workflows: Prepare publish-ready preview assets without opening an editor.
Thumbnail facts
A few numbers worth knowing before you choose a size or a format.
Key points
- 1280×720 is YouTube's stated thumbnail size and the most common 16:9 target.
- 1080×1920 is the vertical cover size used by Shorts, Reels and TikTok.
- 1200×630 is the link preview size read by most social platforms and messaging apps.
- High contrast frames are easier to read at small sizes than detailed ones.
- WebP is typically smaller than JPEG at matching quality, and both are far smaller than PNG for photographic frames.
- Consistent framing across a channel or a course improves how a library reads at a glance.
How it works (technical overview)
Everything happens on your own device. The video is read where it sits, the pictures are decoded by the same hardware that plays video, and the images are written locally — no part of the file is uploaded, and there is no server involved in producing the result.
Steps
- Open: The container is parsed and the device is asked whether it can decode what is inside. Nothing is decoded to answer that.
- Plan: Requested moments are clamped into the clip, rounded, de-duplicated and sorted.
- Decode: One forward pass over the video reads every requested moment, decoding each part of the file at most once.
- Fit: Each picture is cropped or padded into the target box and reduced in successive halving steps.
- Encode: PNG, JPEG or WebP is written, and the result is checked to confirm it really is the format you asked for.
Frames are decoded by your device's own video hardware, one pass per capture set.
Frequently asked questions
How do I extract a thumbnail from a specific timestamp in my video?
Scrub the frame time slider to the exact moment you want — the preview updates live and shows the same crop, size and format the file will have, so what you see is what you download. For single-frame precision, the −1 and +1 buttons step exactly one frame at the source's own rate, which is how you get past a blink or a motion blur. Then press Capture.
Can I generate multiple thumbnails from different parts of the same video?
Yes. Ask for up to fifty captures and each one gets its own slot with its own time, or press Space evenly to spread them across the whole clip in one click. Every requested moment is read in a single pass through the video rather than by seeking to each one separately, so fifty captures cost barely more than five. The whole set downloads as one ZIP.
What output size should I use for a YouTube or social media thumbnail?
YouTube's own thumbnail specification is 1280×720 at 16:9, a Shorts or Reels cover is 1080×1920, and a link preview card is 1200×630 — all four are presets here, so you do not have to remember them. Choose Crop to fill the frame and trim the overflow, or Letterbox to fit the whole frame inside the box and pad the rest. PNG keeps text and edges sharp; WebP is usually the smallest at the same quality.
Why does my extracted thumbnail look blurry or low quality?
Output sharpness is limited by the source: a 720p video does not contain the detail for a sharper image than 720p. So a size larger than the source is treated as a ceiling and clamped to the frame's real dimensions rather than enlarged, and the rail tells you when that has happened. If you genuinely need the larger canvas — a platform that rejects anything smaller — turn on Allow enlarging and the exact size you asked for is produced.
What is a contact sheet, and when would I use one?
A contact sheet is every capture tiled into a single image — the storyboard grid you see in video players and media catalogues. It is the fastest way to review a long recording at a glance or to hand someone a summary of a clip without sending the clip. Choose the number of captures, switch the sheet on, and set the columns and the spacing between cells.
Which video formats can I use, and what happens to the ones you do not support?
MP4, MOV, WebM, MKV and animated GIF open here, and what plays inside them depends on the codecs your own device can decode. Anything else — AVI, WMV, FLV, MPEG, 3GP or OGV — is refused the moment you pick it, with a link to the converter, rather than accepted into a progress bar that fails a minute later. Nothing is uploaded at any point.
from 1,285 ratings
Rate this tool
Tap a star — it takes a second