Video-Based Codec Optimization
Effective Video Transcoding Project (in cooperation with CS MSU Graphics & Media Lab)
Basic info about Effective Video Transcoding
For almost 14 years, Lomonosov MSU Graphics&Media Lab’s video group has been conducting video codecs comparisons. We know that almost always there is a possibility to find efficient encoding options for every video. We did a great study and learned how to determine optimal codec settings for a large number of video classes. After comparing our encoding options for x264 to the 2015 and 2016 comparisons’ options and standard ones, we received very interesting and promising results.
15% bitrate savings
Encoding presets determined by our method
beats x264 developers' presets with keeping
encoding time and encoded video quality
These results are valuable for options giving the same or better objective quality of encoded video and the same or higher encoding speed comparing to standard presets. Thus, we can provide optimal encoding settings that will be a part of pareto optimal RD-curve, that is “shifted” from given encoder options.
We developed a way to find optimal encoding settings for a large number of video classes
Percentage of file size reduction in average for a representative dataset of 77 videos:
You use standard presets and don’t believe that it will work for your videos? Give us a chance — request a demo, for free!
- Give us your video and preset
Send us uncompressed video and encoding settings that you were going to use for compression
- Receive a report
with optimal presets for your video and their gain under your preset
- Choose and pay
We offer additional options for better compression and analysis
- Get encoding settings
and encode similar videos with it
Get your video
compressed with chosen preset
A more detailed technical report and examples will be available soon in this section. At the moment we can share a couple of cool charts.
We created a representative dataset of 385 videos chosen from 9000+ FullHD&4K videos
12 million encoder launches were done on Intel Xeon E3-1125v3
Project mentor: Dr. Dmitriy Vatolin
Thanks to UMNIK and Eureka!Concept programs
- Codecs Comparison & Optimization
- Video Filters
Semiautomatic Visual-Attention Modeling