# Mac Video Compression Test: Picmal, HandBrake, macOS > I ran 3 clips through Picmal, HandBrake and macOS 5 times each, and scored size, speed and quality. Every number is here, including where Picmal loses. Source: https://picmal.app/blog/mac-video-compression-test Picmal's High quality preset compressed my 3 test clips 2.6 to 3.4 times faster than HandBrake's default, with higher quality scores. File sizes were close: Picmal won on one clip and HandBrake on two. Picmal's default preset, Balanced, made the smallest files and had the lowest scores. I make Picmal, so read this with that in mind. Every number below is what the test measured, and the method is at the end so you can check it. ## The results Each clip is 40 seconds long. Quality is the VMAF score, where a higher number is closer to the source. Time is the median of 5 runs. | Clip | App and preset | Size | Saved | VMAF | Time | | :-- | :-- | --: | --: | --: | --: | | Sintel, action (61.5 MB) | Picmal Balanced | 10.3 MB | 83% | 87.2 | 4.5 s | | | HandBrake Fast 1080p30 | 24.3 MB | 60% | 93.2 | 15.8 s | | | Picmal High quality | 30.0 MB | 51% | 95.6 | 4.6 s | | | macOS HEVC 1080p | 35.2 MB | 43% | 97.2 | 6.0 s | | | Picmal Maximum quality | 61.3 MB | 0% | 97.9 | 4.5 s | | Sintel, dark (53.8 MB) | Picmal Balanced | 3.5 MB | 93% | 87.7 | 4.8 s | | | Picmal High quality | 10.7 MB | 80% | 94.3 | 4.8 s | | | HandBrake Fast 1080p30 | 12.3 MB | 77% | 93.9 | 14.4 s | | | Picmal Maximum quality | 23.4 MB | 57% | 96.8 | 4.8 s | | | macOS HEVC 1080p | 33.7 MB | 37% | 96.9 | 6.0 s | | Tears of Steel (20.2 MB) | Picmal Balanced | 4.2 MB | 79% | 88.4 | 2.4 s | | | HandBrake Fast 1080p30 | 8.0 MB | 60% | 94.0 | 6.2 s | | | Picmal High quality | 10.0 MB | 51% | 96.0 | 2.4 s | | | Picmal Maximum quality | 19.8 MB | 2% | 98.2 | 2.4 s | | | macOS HEVC 1080p | 27.2 MB | -35% | 97.6 | 4.0 s | ## Where Picmal loses HandBrake made smaller files than Picmal High quality on 2 of the 3 clips, by about 20%. Its scores were 0.4 to 2.4 points lower on all 3, so you trade a little quality for that size. If you can wait a few more seconds per clip and want every megabyte back, HandBrake is a very good deal. Balanced is Picmal's default, and I kept it that way on purpose. It's for files you want to send or store small, where a file under half the size of High quality's matters more than a perfect copy. If you plan to edit the video again or keep it as your master, use High quality. Maximum quality saved almost nothing on these clips: 0%, 2% and 57%. They were already compressed, which is probably why that preset had little to remove. ## Where macOS fits The macOS encoder scored 96.9 to 97.6, and it took 1.5 to 2 seconds longer than Picmal on each clip. Its files stayed large whatever went in, 27 to 35 MB, and on Tears of Steel that was 35% bigger than the source. For a file that's already small, it can make things worse. ## What this test doesn't cover It's a small test. All 3 clips are 24 fps film, already compressed, and none is 4K. There's no phone video or screen recording here, and those compress differently. Permute and Compresto aren't in the table (here is [Picmal against Permute](/blog/permute-vs-picmal)). Treat this as 3 samples, and test your own files before you decide. The [compress video page](/compress-video-mac) shows what Picmal does with yours, and the trial runs for 7 days. If you're choosing between the two apps, the [HandBrake alternative page](/alternative/handbrake) compares the rest. ## How I ran it I used an Apple M3 Mac with 16 GB of memory on macOS 27.0.1, with other apps open. Picmal 2.1.1 ran through its command line tool, HandBrake 1.11.2 through HandBrakeCLI, and macOS through Apple's `avconvert` with its HEVC 1080p preset. Each used the named preset and no other setting. The clips are 40 seconds cut without re-encoding from [Sintel](https://durian.blender.org/) (1080p) and [Tears of Steel](https://mango.blender.org/) (720p), both by the Blender Foundation under CC BY. I ran 5 rounds for each app and clip. HandBrake's times moved the most, from 13.7 to 17.4 seconds on the action clip. Every output kept the source's frame count, and I scored each one against its source with VMAF in FFmpeg. Last updated: .