Frame Extraction: FPS vs Every-N-Seconds — and How Browser-Based Extraction Actually Works

You want to pull stills out of a video. Every extraction tool asks you the same question: how many frames do you want? Usually that arrives as two options — "extract at N frames per second" or "extract one frame every N seconds." They sound like two ways of saying the same thing, and for many videos they are. But pick the wrong one and you end up with 3,600 near-identical images of a talking head, or a sparse sample that skips the one moment you needed.

This guide covers both decisions at once: how browser-based frame extraction actually works under the hood (the part competing tools rarely explain), and how to choose between FPS mode and interval mode so you get exactly the frames you need — no upload, no server, nothing leaving your device.

How a Browser Extracts Frames (the "Extraction Engine")

Before the choice of cadence makes sense, it helps to know what the tool is physically doing, because "extracting a frame" is not like copying a file. A video file does not contain 30 pictures per second sitting in a folder. It contains a compressed stream: a handful of full reference images (keyframes) plus mathematical instructions for reconstructing everything in between. The only thing that can turn that stream back into pictures is a decoder.

A browser-based extractor works like this:

1. Your video is loaded into a local <video> element — the same element any web page uses to play video. No upload happens; the file stays on your machine. The browser's native media decoder (the same one that renders Netflix-quality playback) does the decompressing.
2. The engine jumps to each requested timestamp by setting the video's playback position and waiting for the seeked event — the browser's signal that the decoder has reconstructed that exact moment.
3. The frame is painted onto a <canvas> with drawImage, which snapshots the decoded picture pixel-for-pixel at the video's real resolution.
4. The canvas is exported as JPG, PNG, WebP, or AVIF and packaged for download — on this site, as individually numbered files or one ZIP archive with zero-padded sequential names (frame_0001.jpg, frame_0002.jpg, …) so they sort and import as an ordered image sequence in any editor.

Two properties of this architecture matter for everything below. First, quality is bounded by the source: the frame you get is the same picture the browser would display at that timestamp — no better, no worse. Extraction never "enhances" a frame; it snapshots what the decoder decoded. Second, frames are produced by seeking, not by playing the video in real time. The engine asks for timestamp 1.25s, waits for the decoder, captures, then asks for 2.5s. That means extraction speed depends on how fast the browser can seek and decode — which is why a 10-minute 4K video takes noticeably longer than a 1-minute 720p clip, even though both are "just extracting frames."

This seek-based design is also why tools that run in the browser can do the job with no server: everything the decoder needs is already in your browser, and everything the exporter produces never leaves your tab. Some competitors market this as "zero-RAM streaming" or similar; the honest version is simpler — the work happens locally because the browser already contains a professional-grade video decoder.

Sources: MDN — HTMLVideoElement, MDN — CanvasRenderingContext2D.drawImage

The Two Cadences: FPS Mode vs Every-N-Seconds Mode

Given the seek-based engine, both extraction modes are just different ways of generating the list of timestamps to visit:

• FPS mode ("N frames per second"): the engine divides the video into evenly spaced slices — one frame every 1/N seconds. At 1 fps, a 10-minute video yields 600 frames. At 0.5 fps, 300 frames. (Fractional FPS like 0.2 = one frame every 5 seconds is the same math.)
• Interval mode ("one frame every N seconds"): you state the gap directly — one frame every 5 seconds, every 30 seconds, every minute. It's the same arithmetic viewed from the other end.

Mathematically they're interchangeable: 1 frame every 5 seconds is 0.2 fps. The difference is which one is easier to reason about for your task, and a practical one in the UI — interval mode tends to come with presets tuned for review workflows ("every 10s", "every 30s"), while FPS mode suits frame-accurate production work where you think in frame rates.

The real decision isn't FPS vs interval. It's how many frames you actually need, because the count is what costs you time (extraction), bandwidth (download), and disk space. The formula is unavoidable:

A 5-minute video at 1 fps = 300 frames. At one frame every 10 seconds = 30 frames. Same video, same tool, 10× difference in output. Before touching any setting, do that multiplication — it's the single highest-leverage thing in this guide.

When to Use Each: A Practical Decision Guide

The Privacy Part Nobody Reads (But Should)

Here's the part that distinguishes the architecture above from server-based tools: the video never leaves your device. There is no upload progress bar because there is no upload. Your file — a client contract draft, unreleased footage, a family video — is decoded in your browser tab and the extracted images are generated there too.

That isn't just a privacy nicety; it's a throughput property. No upload means no waiting on your connection, no queue behind other users' jobs, and no terms-of-service question about who owns the file while it sits on someone's server. When competitors advertise "no upload" as a feature, this is what they're describing — the same local-decoder architecture, packaged as a trust signal. Verify it yourself: start an extraction and watch your network tab. Nothing moves.

One honest limitation of the local approach: DRM-protected video can't be extracted this way. If the browser refuses to expose a video's pixels (encrypted streams from paid streaming services), a canvas snapshot will come back blank — by design, enforced by the browser. That's not a bug in any extraction tool; it's the platform working as intended. For anything you recorded, downloaded legitimately, or rendered yourself, this doesn't apply.

A 60-Second Workflow

1. Check the source: open the file in Video Metadata Inspector — note duration, resolution, codec.
2. Do the math: duration × your FPS (or duration ÷ interval) = expected frame count. If it's in the thousands, widen the interval.
3. Pick the mode: density → FPS; coverage → interval; uneven pacing → smart scene detection.
4. Extract at Video to Frames, download the numbered ZIP, and import the sequence straight into your editor.

Frequently Asked Questions