# In-Browser Processing Benchmarks — Instant vs Seconds

> Measured timings for in-browser processing: which operations are truly instant (lossless stream copy), which take seconds, and why. From our own tests.

Source: https://toolpantry.app/local-processing-benchmarks

## About

Browser-based processing has two tiers: operations that are truly instant (and provably lossless), and operations that re-encode and take seconds. We measured both.

## Quick answer

The short version: trimming, muting, rotating, and remuxing a video finish in 0.0 seconds — they copy the original data instead of re-encoding it. Compressing or blurring video takes seconds to tens of seconds , scaling with clip length and your device's hardware encoder.

## Methodology

| Operation | Test input | Time | Fidelity check |
|---|---|---|---|
| Trim video ([video trimmer](https://toolpantry.app/video-trimmer.html)) | 6 s 720p MP4, cut to 4 s | 0.0 s | Video stream bit-identical |
| Remove audio ([mute video](https://toolpantry.app/mute-video.html)) | Same clip, audio dropped | 0.0 s | Video bitstream MD5-identical |
| Rotate 90° ([rotate video](https://toolpantry.app/rotate-video.html)) | Same clip, metadata patch | instant | Zero pixel difference vs reference |
| MOV → MP4 ([remux](https://toolpantry.app/mov-to-mp4.html)) | Same clip in .mov wrapper | 0.0 s | Video + audio bitstreams identical |
| Rotate PDF pages ([pdf rotate](https://toolpantry.app/pdf-rotate.html)) | 3-page PDF | instant | Page attribute only; content untouched |

These operations never decode a single frame. Trimming copies the video samples between two keyframes; muting leaves the audio track out; rotation writes a different display transform into the file header; remuxing rewrites the container index. That is why they take the same time for a 10-minute file as for a 10-second one — the limit is disk copy speed, not processing.

| Operation | Test input | Time (software mode) |
|---|---|---|
| Compress image ([image compressor](https://toolpantry.app/compress-image.html)) | 1.13 MB photo | < 1 s |
| Trim audio / export MP3 ([audio trimmer](https://toolpantry.app/audio-trimmer.html)) | 3.5 s MP3 | < 1 s |
| Blur faces in video ([video face blur](https://toolpantry.app/face-blur-video.html)) | 6 s 720p clip | ~15 s |
| Compress video ([video compressor](https://toolpantry.app/compress-video.html)) | 1 min 1080p clip | 10–40 s (typical laptop) |
| Merge two clips ([merge video](https://toolpantry.app/merge-video.html)) | 2 × 6 s 720p → 12.1 s output | ~49 s |

Re-encoding decodes every frame, draws or scales it, and encodes a new stream — the one quality generation that compression, blur, crop, and merge inherently cost. Two levers make it faster: a shorter clip (linear), and hardware acceleration (a real browser with GPU access, unlike our test harness).

## Why this matters for choosing a tool

If your task is on the instant tier — cutting, muting, rotating, converting containers — a browser tool isn't just more private than a cloud service; it's *faster*, because there's no upload. If your task requires re-encoding, the browser is still private and often competitive, but the honest advice is: check the result, and keep the original. Our [compression benchmark page](https://toolpantry.app/compression-benchmarks.html) has the quality-side numbers; the [file size limits reference](https://toolpantry.app/file-size-limits.html) covers where the results need to land.

## FAQ

**Why are some file operations instant and others take time?**

There are two classes of operation. Lossless ones — trimming, muting, rotating, or remuxing a video — rearrange or re-label existing data without decoding it, so they finish as fast as the file can be copied (often under a second). Re-encoding operations — compressing video, blurring faces — must decode every frame and encode a new one, which takes seconds and scales with clip length.

**Is the "instant" tier really lossless?**

Yes, and we verify it. In our tests, a muted video's bitstream was MD5-identical to the source, a rotated video's frames matched a reference rotation with zero pixel difference, and a MOV-to-MP4 conversion produced identical video and audio bitstreams. The original data is copied, not re-created.

**Does processing speed depend on my device?**

Yes. Re-encoding uses your device's hardware encoder when available, so a modern phone or laptop can be several times faster than our headless test environment, which runs in software. Lossless operations depend mostly on disk read/write speed and are fast on any device.

**How long does a typical video compression take?**

On a normal laptop, a 1-minute 1080p clip typically takes 10–40 seconds to compress. Longer clips scale roughly linearly. In our software-only test environment, merging two 6-second 720p clips took about 49 seconds — hardware acceleration in a regular browser is significantly faster.

**Why process files in the browser at all?**

Two reasons: privacy and time. Nothing is uploaded, so sensitive files never leave your device — and you skip the upload/download round trip entirely. For lossless operations the browser is actually faster than any cloud tool, because there is nothing to transfer.