nablet News

mediaEngine. 50% faster Video Transcoding.

nablet Team made a comparison of nablet's mediaEngine and FFmpeg to measure the overall transcoding speed and visual quality of transcoded material.
To compare the products, we run the same tests on a laptop and a standard PC running Windows 10 to check different hardware configurations. To make the comparison fair, we used the same source content for both engines, and we tried our best to make the transcoding configuration similar. It means that the target bitrate, keyframe interval, and encoding levels are the same, and the presets are similar. 

As content, we used different materials:
  • a sea shore with a lot of water to make the encoding complex, 
  • highlights of a football match with a lot of movements,
  • Studio test footage.
The content is of different frame rates and resolutions. And we didn't use any video conversion operations - neither frame rate, nor interlacing, nor scaling. The goal was to measure a pure transcoding performance.

We run tests to compare the performance of the most popular video codecs: AVC, XDCAM, and ProRes. 

The main result is that mediaEngine runs at least 25% faster than FFmpeg. In some scenarios, the results are 50% faster and even higher. The visual quality is about the same. On low bitrates, the difference is about 2-3%. And on high bitrates, there is no visual difference. To measure the visual quality, we used PSNR, SSIM, and VMAF metrics. And the measurement of the processing time and frame rate was done by built-in features of mediaEngine and FFmpeg. More details about the results you can find in a whitepaper

The reason for such a big performance gain of mediaEngine is in its highly-optimized architecture. nablet development team puts a lot of effort in getting the best performance results while keeping the video quality as good as possible on any hardware. Together with 10+ years of expertise in video codecs development, it makes mediaEngine a fast and reliable video transcoding factory for your media processing scenarios.