Skip to content
AbeyIO

04 / Choose a local method

How to extract video frames without uploading

Keep the video on your device. Choose a browser tool for a few stills, local FFmpeg for repeatable commands or every decoded frame, or a screenshot for a quick visual note.

Quick answer: for separate images without an upload or account, open AbeyIO Frame Extractor, choose a local video, select a timestamp or interval, and download a PNG, JPG, or WebP ZIP. For every decoded frame or command-line automation, use an installed copy of FFmpeg with a local input path. For one grid instead of separate files, use Contact Sheet.

Browser tool, FFmpeg, or screenshot?

The right method depends on the output. A frame export saves decoded video imagery. A screenshot captures the displayed player, which can include controls, scaling, borders, and other screen content.

Choose by task and limits, not a ranking
MethodUse it forOutput and limits
AbeyIO browser toolOne still or a regular series without a command-line setupZIP of PNG, JPG, or WebP. Up to 1 GB input and 600 interval samples. Timestamp and interval modes need browser-playable SDR video.
Local FFmpegRepeatable commands, larger batches, or every decoded frameNumbered image files. Requires installation, a supported decoder, and enough disk space. HDR needs a separate color workflow.
Manual screenshotOne quick note about what a local player displaysA screen capture, not necessarily the video's pixel dimensions. Crop player controls and check display scaling.
AbeyIO Contact SheetOne visual index of the clipOne PNG or JPEG grid with 12, 24, or 48 samples. Not a set of full-size images.

If the question is separate images versus one grid, read the Frame Extractor versus Contact Sheet comparison. If a brief event must be found, review the video or decoded frames around it; sparse samples can miss it.

How do I save one photo or a series in the browser?

  1. Open Frame Extractor and select the video from your device. Start with an H.264 MP4 that the browser can play.
  2. For one image, open timestamp mode. Enter a position in seconds before the video ends.
  3. For regular samples, open interval mode. Use 5 for a five-second interval. Set a start and optional end; the end is excluded.
  4. Open Options. Choose PNG, JPEG, or WebP. Set Max width to Original to avoid resizing.
  5. Select Extract Frames ZIP. Unzip it to get the images and manifest.json. Even single-image mode downloads a ZIP.

The manifest records requested timestamps, not verified source-frame times. Browser seeking selects the displayed frame. For counts and time ranges, use the interval sampling guide.

Need every frame? Do not use interval sampling or duplicate removal for that requirement. Duplicate removal can discard matching images even at threshold 0. The FFmpeg method below avoids those two filters.

How do I extract frames with local FFmpeg?

These commands need FFmpeg installed and available as ffmpeg in your terminal. Download the sample-screen-recording.mp4 and run the commands in a new output folder that contains that file. Each input is a local file path, not a URL. The commands leave the input unchanged and use different output names.

One PNG at five seconds

ffmpeg -n -i "sample-screen-recording.mp4" -map 0:v:0 -ss 5 -frames:v 1 -update 1 "still.png"

-map 0:v:0 selects the first video stream. -ss 5 after the input decodes and skips to the requested position. -frames:v 1 writes one image. -n prevents replacing an existing output file.

One PNG per five seconds

ffmpeg -n -i "sample-screen-recording.mp4" -map 0:v:0 -vf "fps=1/5" "sample-%03d.png"

The fps filter produces a regular output rate of one image per five seconds. %03d creates names such as sample-001.png. This filter uses its own timestamp rounding and can drop or duplicate frames. Its samples need not match browser requests at 0, 5, 10, and 15 seconds.

Every decoded frame as a JPG

ffmpeg -n -i "sample-screen-recording.mp4" -map 0:v:0 -fps_mode passthrough -q:v 2 "all-%04d.jpg"

This command has no interval or duplicate-removal filter. -fps_mode passthrough avoids frame-rate conversion at the output. -q:v 2 sets JPEG quality; JPG is still lossy. For PNG, change the output to all-%04d.png and remove -q:v 2. Large clips can produce many files and use substantial disk space.

For variable-frame-rate footage, frame count is not simply duration multiplied by an advertised FPS. Decoder errors, stream choice, and timestamps affect results. These commands do not preserve a forensic chain of evidence or provide an HDR-to-SDR tone-mapping recipe.

Worked example: one 20-second recording

On 29 October 2026, we ran the three commands above with FFmpeg 8.1.2 on macOS. The linked input is a 20-second, 1280 × 720 H.264 MP4. FFprobe reported 600 video frames at 30 fps. We counted the output image sequences with FFprobe.

Measured command outputs for the linked sample only
CommandMeasured outputImage dimensions
One still at 5 seconds1 PNG1280 × 720
fps=1/54 PNGs1280 × 720
Every decoded frame600 JPGs1280 × 720

These counts are measured results for this clip, not performance benchmarks or promises for other files. For the browser, interval 5 with start 0 and end 20 requests four images at 0, 5, 10, and 15 seconds. That is a calculated browser example, not proof that the two methods select identical frames.

Preview of separate frames from the sample release checklist recording
Existing browser-extraction preview from the same source recording, not output from the FFmpeg commands above. Download the browser example ZIP.

How do I keep the original image detail?

Export at the video's decoded pixel dimensions and avoid an extra lossy image-encoding step when detail matters. In AbeyIO, use PNG and Max width: Original. JPEG can be useful for smaller files, but repeated lossy encoding can remove detail. Neither format restores information missing from the video.

A 1280 × 720 video does not become a true 4K image when enlarged. A screen capture may use a different pixel size from the source. Check the source resolution and codec before choosing output settings. HDR color conversion and color management are separate from image dimensions.

Does no upload mean completely offline?

No. AbeyIO processes the selected video locally, but page assets, analytics, error reporting, and optional feedback use network requests. Some duplicate-removal paths need a decoder download on first use. An unused page is not guaranteed to open offline. See the processing and privacy limits.

The FFmpeg commands here use a local input path and local outputs. FFmpeg can also accept network URLs, so that distinction depends on the command you run. If you need to work without a connection from the start, prepare the installed tool and input files before disconnecting.

Technical references: FFmpeg's command options, fps filter, and image-sequence output; MDN's browser seeking reference. The worked commands were checked against the installed FFmpeg help and run on the linked clip.