Video Processing Speed
When processing long and/or high resolution videos, you will probably ask yourself “Why the hell is it taking so long?!”
This is actually a good question and to make this issue more understandable – particularly for those new to video editing – we decided to do some benchmarking that shows which aspect of video editing has which effect on processing times.
The first two aspects are rather easy to grasp:
- Your computer’s processing power
- The amount of video data to process: The longer the video’s length, the higher the resolution and the more fps (frames per second) your video has, the longer it will take.
The third aspect is a bit harder to pinpoint but is just as important:
- How much processing are you applying? Whether you just trim and re-encode your video or apply color and contrast correction, fisheye removal and other filters will obviously have a big impact, simply because there’s a lot more stuff to compute.
How We Tested
To give you a better idea of how these factors work we used Video Improve v2.6.6 and ran benchmarks on three different computers:
- PC 1: Windows 10 x64 Desktop, AMD Ryzen 7 1800X (8 cores, 3.6GHz), 32GB RAM, GeForce GTX 1080*
- PC 2: Windows 10 x64 Laptop, Intel Core i7-7700 (4 cores, 2.8GHz), 16GB RAM, GeForece GTX 1060*
- Mac 1: MacBook Pro (Mid 2014), macOS 10.14.6 , Intel Core i5 (2 cores, 2.6GHz), 8GB RAM, Intel Iris*
* Though listed here, the GPU is rather irrelevant because we currently use hardware acceleration for video decoding only, which has little impact.
Also, while the Mac sports much weaker hardware than the two Windows machines, please don’t see this as one of those pointless “PC vs Mac” discussions. These are simply the machines that were readily available for testing.
The benchmark consisted in processing 3 videos (all of them being 5 minutes long, 30fps, encoded with H.264) in 3 different resolutions:
- 1280x720 (HD)
- 1920x1080 (Full HD)
- 3840x2140 (4K)
For each video, we ran three different processing scenerios:
- Re-code only
- Remove fisheye distortion (we used the GoPro7 video lens profile)
- Remove fisheye distortion and apply the “Paris” preset from the Liquivid Preset Library, which adds extensive adjustments to exposure, colors, filters, sharpness and even a tilt-shift effect.
That makes 27 processing runs (3 computers * 3 videos * 3 processing setups).
|Re-code only||PC 1||PC 2||Mac 1|
This already offers some insights:
- Going from 1280x720 to 1920x1080 will make encoding roughly twice as slow, and going from 1920x1080 to 4K will make it roughly another 4 times slower. This is not at all surprising, as this reflects the increase in image pixels from one format to another:
- 1280x720 = 0.9 megapixels
- 1920x1080 = 2.1 megapixels
- 3840x2160 = 8.3 megapixels (per frame)
- The used hardware has a huge impact, with the fastest computer being roughly 10 times faster than the slowest. Note that the (fast) PC 1 is not the fastest thing that money can buy at the moment and that the (slow) Mac 1 is still a solid machine for surfing and other office tasks.
And this is the catch: Only because your computer may feel fast enough when surfing and writing emails does not mean it’s anywhere near fast enough for video editing.
|Remove fisheye distortion||PC 1||PC 2||Mac 1|
While the ratio between the individual processing times remains rather similar, hardware limitations are now becoming more prominent, with the slowest machine now taking amounts of time that even patient users will probably not want to spend.
|Remove fisheye distortion|
+ “Paris” preset
|PC 1||PC 2||Mac 1|
In our most intense processing setup you can see times increasing a lot – there’s simply a lot of stuff to compute in order to apply all of the adjustments and filters. While it’s slow even on our fastest machine, you can pretty much forget about doing this on a weak computer for anything larger than 1280x720 and even that would test your patience (remember: this is for a 5 minute, 30fps video).
Hardware Encoding (macOS only)
With Video Improve v2.7 we have added support for hardware-accelerated encoding on macOS. This means that unlike the classic (software) encoder, the GPU (or other specialized hardware) can be used to speed up video encoding (for the H.264 codec only).
|Re-code only||Mac 1 (software encoding)||Mac 1 (hardware encoding)|
This is a significant improvement, with encoding now being between 2.4 and 3.6 times faster. Unfortunately, the hardware encoder produces somewhat larger files at somewhat lower visual quality but taking into account how much time it saves you, you should definitely consider using it in most cases.
Sadly, we currently don’t support hardware encoding on Windows.
Yes, video processing can be slow.
Particularly if you have an old machine or have 4K footage or both. This can be particularly painful if you just bought a new GoPro or even one of those more affordable action cams and are happy about finally being able to do 4K just to find out that your computer won’t suffice. While 4K cameras can be quite cheap, a computer capable of handling those videos probably isn’t so think well whether you actually need 4K.