How do I extract frames from video for a dataset?

Sample on a fixed interval (often 1–5 fps), export lossless PNG when labels need clean edges, and keep filenames stable. GetVideoFrames does that locally so proprietary footage never uploads to a converter.

What does video to frames 1fps mean in practice?

One frame per second of video. A 10-minute clip yields about 600 images — usually enough for coverage without the storage cost of every frame at 30 fps (18,000 images).

Should ML datasets use PNG or JPG?

PNG is safer for annotation and augmentation pipelines that re-encode often. JPG is fine for large web-scale sets where storage dominates. This tool supports both plus ZIP packaging.

Do you include timestamp naming and a manifest?

Yes. Dataset mode favors consistent frame names and ships a manifest.json in the ZIP so loaders can map index, timing, and file paths without hand-written scripts.

When should I use FFmpeg instead?

Use FFmpeg for headless servers, CI jobs, and multi-file shell pipelines. Use GetVideoFrames when a researcher needs a private, visual pass on a laptop today. Our FFmpeg guide covers the CLI path in depth.

1–5 fps · Manifest · Private

Extract Video Frames for Machine Learning Datasets

Build video to frames 1fps–5fps datasets in the browser. Free local export with timestamp-friendly naming and manifest.json — no upload.

The extraction controls are currently available in English.

Local frame workspace

Drop a video here

MP4, MOV, WebM, MKV, AVI — up to 2 GB. Processed on your device. Nothing uploads.

MP4MOVWEBMMKVAVI

Drag, browse, or paste a video

Why sampling beats dumping every frame

Most vision tasks do not need 30 near-duplicate images per second. Sampling at 1–5 fps cuts storage, labeling cost, and redundancy while keeping temporal coverage. GetVideoFrames is built for that extract-frames-from-video-for-dataset pattern without sending footage to a third-party host.

Privacy for sensitive training footage

Medical, industrial, and internal product videos often cannot leave the laptop. Local extraction means you can still prepare frames for labeling tools while satisfying data-handling rules. Upload converters that keep files for ~1 hour are a non-starter for those teams.

PNG lossless, naming, and manifest.json

Prefer PNG when bounding boxes and masks must stay crisp. Keep sequential or timestamp-oriented names so train/val splits stay reproducible. The ZIP includes a manifest.json describing the exported set — a lightweight bridge before full COCO/YOLO packaging arrives in Pro.

FFmpeg alternative for automation

For servers and scripts, FFmpeg remains the standard. Use fps=1 or fps=1/5 in a filter graph when you need unattended jobs. When you want a visual, private pass first, stay here — then graduate to CLI with our complete FFmpeg frame guide.

Related format choices

Need lossless editing stills without dataset naming? Try video to PNG. Need lighter preview frames? Use video to JPG.

FAQ

Frequently asked questions

How do I extract frames from video for a dataset?: Sample on a fixed interval (often 1–5 fps), export lossless PNG when labels need clean edges, and keep filenames stable. GetVideoFrames does that locally so proprietary footage never uploads to a converter.
What does video to frames 1fps mean in practice?: One frame per second of video. A 10-minute clip yields about 600 images — usually enough for coverage without the storage cost of every frame at 30 fps (18,000 images).
Should ML datasets use PNG or JPG?: PNG is safer for annotation and augmentation pipelines that re-encode often. JPG is fine for large web-scale sets where storage dominates. This tool supports both plus ZIP packaging.
Do you include timestamp naming and a manifest?: Yes. Dataset mode favors consistent frame names and ships a manifest.json in the ZIP so loaders can map index, timing, and file paths without hand-written scripts.
When should I use FFmpeg instead?: Use FFmpeg for headless servers, CI jobs, and multi-file shell pipelines. Use GetVideoFrames when a researcher needs a private, visual pass on a laptop today. Our FFmpeg guide covers the CLI path in depth.