Product Development

advanced compression for live streaming/video transfer

Ram malay CEO - Soniks Consulting, Founder & CEO www.superbigstore.com

June 23rd, 2020

Topic: advanced compression for live streaming/video transfer.

We have developed product "webkonf meetings" we are in process of fine tuning certain areas to make it more robust. like advanced compression techniques on live streaming and video calls. Anyone worked on these extensive algorithms and ready to share inputs?

Just FYI: we are already using open source compression techniques. Looking for something advanced and might not be used or known outside. We are ready to use it and test it before using in live instance.


Thanks,

Ram Malay

Founder & CEO - webkonf.io

linkedIn: https://www.linkedin.com/in/rammalay/

www.webkonf.io

GK CTO / Consultant - AI, Blockchain, Analytics

Last updated on June 24th, 2020

Compression is context-sensitive and the algorithm is to be picked based on the target that you are trying to optimize: For example, latency vs throughput.

From the question, it is not clear which context you are operating in (such as bandwidth-constrained areas, or targeting high-quality image, for example in medical domain tele-health).

If it is just a regular web-conference, akin to webex etc. then any lossy codec that uses localized temporal differential encoding should suffice, with motion-compensation added in few cases (e.g. WebRTC already supports H264 and should be enough).

Overall, if you are using open-source, then it is suggested to go with existing techniques since it is a solved problem. If you really want to create ground-breaking encoding, then you can consider object-based feature compression models. This would require utilizing transfer-learning methods (of AI) to encode the inter-frame motion and can drastically reduce the bandwidth requirements.

Also in some cases, more than the compression, the backend infrastructure is going to play a more important role. For example, are we talking about centralized session relays using media server streams, or are we talking about P2P relays?


-GK

AShu Co-founder I CTO I Co-adviser I Full stack developer

Last updated on June 24th, 2020

Hey Ram

I am also working on project like you .

For now as my experience use H.265 or HEVC this is the future of video encoders

this page but HEVC need more cpu than AVC and performances of HEVC more better than AVC.

for more info see

https://aws.amazon.com/media/tech/high-efficiency-video-coding/


And for live streaming HLS is best way for lives streaming videos

you can use aws transcorder and cdn for streaming video or simulcasts

Thanks

Ram malay CEO - Soniks Consulting, Founder & CEO www.superbigstore.com

June 24th, 2020

we are using webRTC. compression technique is H264. But we are exploring alternatively if we can find better compression techniques so that our bandwidth compression would reduce and performance/user experience would increase compared to competetive products.

We are exploring AI/ML but it takes time. So we are exploring if anyone is already working on this and ready to share AI algorithms with us to test and use it.


Thanks,

Ram

Emily Founder & CEO of Greengage. Previously @TheEconomist @EliteDaily @BCG

June 28th, 2020

My business Greengage Ventures uses a lot of livestreaming (as well as recorded video) for theatre and music performances with global audiences. Very happy to test your product or provide an industry perspective for your product needs etc