Automatic Video Compression: Pick Quality, Not Settings

Automatic video compression means the software works out every technical setting for you, file by file. You say how you want the result to look, in plain words. It does the engineering.

That sounds like the obvious way to build such a tool. It is also strangely rare. Most compressors still hand you a wall of knobs and wish you luck. I have not seen another consumer app that probes each file and picks settings for that one file. That gap is a big part of why I built Video Squisher.

Presets make you the engineer

Open most video compressors and the settings wall greets you first. Bitrate. CRF numbers. Preset ladders with names like “slow” and “veryslow”. Resolution dropdowns. Some of it is honest engineering. None of it is your job.

The wall hides a real problem. The right setting for one video is wrong for the next. A phone clip shot in a bright kitchen compresses nothing like a shaky concert video filmed in the dark. Same camera and same resolution, but a completely different file inside. A fixed number cannot know that.

Here is a concrete version of the problem. Two files can both be “a 4K video” and want opposite treatment. One is handheld footage full of swaying grass and moving leaves, which is brutal to compress. The other is clean, steady footage with long still shots, which shrinks easily. Give both the same preset and one gets hammered while the other barely gets smaller.

So the presets quietly push the hard decision back onto you. Pick a bitrate too low and faces turn to mush. Pick one too high and you paid for bytes nobody needed. You came with a simple question and left with homework.

Presets exist because they scale well for the person building the tool. One settings page ships to everyone. Tuning each file individually does not scale for them, even though it is exactly what would help you.

I have spent an embarrassing amount of my life answering forum posts that boil down to “what CRF should I use”. That is a question a person should never have to ask.

What “automatic” should mean

Real automatic compression starts from a different question. Not “which number should this human type in”. Just: how do you want this to look?

Video Squisher asks it with words. The quality levels are Smallest, Smaller, Balanced, Near Original, and Original. Five plain words, and nothing you need to look up first.

The program analyzes your video and automatically picks the best compression settings for that particular video.

That sentence is a whole pipeline wearing one line. The app treats your file as an individual. It runs test compressions on the actual video and scores every candidate against your source using the same perceptual quality metrics streaming companies rely on: VMAF, SSIM, and XPSNR. VMAF is the big one: a quality score built by Netflix engineers, where a higher number means the compressed video looks closer to your original. I explain all three like you are busy in a separate piece: VMAF and SSIM, explained. The strongest compression that still meets the quality you picked wins.

Notice what you were never asked: no numbers and no abbreviations. The word at the quality picker carries the whole decision, and the app turns it into settings that suit this one file.

The per-file part is what separates this from marketing automatic. Plenty of tools call themselves automatic because they run a preset without asking. Presets are one-size-fits-all answers, and your file is not one size. The analysis here is per file, on the actual bytes you gave it, which is the only version of automatic that can honestly keep a promise.

The wording matters, so let me be precise. “Meets the quality you picked” is a measured bar, not a hope. If you chose Near Original, the app checked that the winner really looks near original, on this file, because it scored it. A promise is only worth the check behind it, so the check comes first.

The checks before saving

A compression app should grade its own homework before handing anything over. This one does.

Before anything is saved, the winning candidate is scored against your source one more time. It only ships if its score clears the level you chose. The quality word you picked is a requirement, not a label.

Verification here means comparing the compressed result back against the source it came from, with the metrics doing the looking. That is a fair test, because the source is the only honest reference either of us has.

None of this needs your involvement. You are not asked to compare screenshots or squint at two windows. The checking is the app’s job, done with the same tools the streaming industry trusts, and it happens every single time rather than when somebody remembers.

The top quality levels always keep the full resolution, too. Near Original does not quietly drop you down a resolution and hope you do not look closely. Output is MP4, using AV1 or HEVC, so the result plays on the usual things.

One real run

Concrete numbers help, so here is one. A real run, not a promise: a 4K phone clip came out 97% smaller, and it still scored 97+ on VMAF. A big file became a small one, with the quality measured instead of guessed.

Now the honest caveats, because I would rather tell you than have you find out later. Every file is different, and that 97% is one clip, not a guarantee. Videos that have already been compressed simply have less room to shrink, whatever tool you feed them. That is physics, not laziness.

Also worth saying: smaller is not automatically better. Squeezing a file that already plays fine everywhere is a perfectly good errand when you are short on disk space. For your only copy of something important, maybe the answer is Balanced instead of Smallest. The words exist so you can make that call without a manual.

And if you are curious what the app actually chose, that curiosity is allowed. The point is that you never have to care.

Where the honest wording lives

People search for “compress video without losing quality” and half the results over-promise. The honest version of that phrase is quality-targeted: you name the quality, the app aims for it. The result then gets measured against the source before you ever see it. I gave that framing its own guide, because the wording deserves the space: compress video without losing quality. I also wrote a whole article on why the usual version of that promise falls apart: Why “compress without losing quality” is usually a lie.

This is the idea the whole app stands on, and Video Squisher is where it lives. Everything happens on your own computer, so there are no uploads and no accounts.

Questions people ask

Is this just an FFmpeg wrapper?

Fair question, and the honest answer is partly. The actual encoding work is done by established encoder software, the same proven building blocks the rest of the industry leans on. A wrapper stops there. The difference is everything around the encoding: the per-file analysis, the test compressions, the scoring against your source, and the choice. The encoders are the hands. The analysis is the brain, and the brain is the product.

What about screen recordings, drone clips, or old camera footage?

The analysis is per file, so different content gets different treatment. A screen recording, a drone shot, and a phone clip each get their own test compressions and their own verdict. Content that is already compressed simply has less room to shrink, and the app still finds the room that exists.

Does automatic mean lower quality?

It should mean the opposite. The level you pick is enforced by measurement, not by vibes. The top levels keep full resolution, and nothing gets saved until it has been scored against your source and passed.

If that sounds like your kind of deal, download Video Squisher and feed it something huge. It runs on Windows, Mac and Linux.

K
KarlKarl builds small, honest apps like Video Squisher and Image Squisher.More about Karl
Scroll to Top