September 21, 2017 by Romain Bouqueau

SUMMARY OF IBC 2017

This year was the 50th anniversary of IBC. IBC is in our opinion the most important show of our industry (followed by NAB in Las Vegas which addresses the US market better). However we find it unfortunate that the organizers feel obliged to report an increasing number of attendees every year. The show is getting better but our perception is that the attendance dropped severely in the last 5 years.

OTT low latency

GPAC Licensing Signals and Keepixo at IBC 2017

Latency is compared between the broadcast and OTT workflows. Broadcast is a few seconds, most of which are concentrated in the encoder to get the best encoding quality. OTT latency ranges mainly between 20 and 45 seconds.

OTT low latency is when the OTT and the broadcast latencies meet. We use the ultra-low latency term when the OTT latency is even lower. Of course nothing to see with video-conferencing transport that can't be cached using CDNs (and thus show unrealistic content delivery costs).

Looking for VR

We followed the news on VR along the year. Many of the startups that brought a lot of fresh air to IBC last year have disappeared, pivoted, or are waiting for the right time to launch their products. Of course some bigger actors such as Ericsson or Accedo had some nice demos but off the schedule in our opinion.

While we think VR is going to be successful, we feel like AR is more ready at this point. The availability of SDKs on all the major platforms consolidates our views.

The future will be made of 360 companion videos (not requiring headsets), and a better resolution (8K headset) and smart glasses.

The dream of AI and automatic metadata

Lots of keywords this year to use the advances of Artificial Intelligence to power automatic metadata:

• machine learning, data-driven advertising, AI,
• conversational, facial recognition, speech-to-text, OCR,
• smart cloud-based microservices,
• massively enriched metadata.

Given the sophistication of the technology above, we are probably 5 years away to get some on-the-shelf products. The demo we saw had too many false-positive to remove entirely human decisions from the process. We did enjoy the vision of Nokia for 2025. It was very interesting to articulate such a vision and the discussions on the booth were interesting.

For automated metadata, we feel like the marketing departments put their best efforts to hide the real points:

• automatically generated metadata on a video automatizes work done by humans ;
• acquiring user data. User data management is becoming a key factor as it allows to create personalized experiences (to be monetized with targeted ads).

Not a word about user privacy on any booth.

HDR: could do better

HDR is experiencing a format war. We heard many rumors of actors buying and willingly segmenting the market. In term of experience we felt that HDR10 is sufficient in most environments. We could see a difference with dynamic metadata but couldn't tell which stream was the best.

In our opinion, the great part of HDR is the fidelity. HDR is the HiFi of video. The marketing talks missed that.

The new HEIF image format can use HEVC to encode still and GIF-likes images. Apple adopted it on iOS 11. Camera vendors wonder if they could not replace JPG (BT.709) exports by HEIF (BT.2020).

IP and software workflows

It is obvious to many of us that software is eating the world. Our OSS fellows at Open Broadcast Systems show the way in this direction.

We also saw some full IP studios (BBC R&D team), workflows, and remote production systems (like the Sony Virtual Switcher that should be available at the end of 2017). More to see in the following years. As more and more people produce video, the intrinsic value of video decreases and the industry needs to low the production costs.

The releases of SMPTe 2110 and AMWA IS-04 are strong points in this direction.

Moving away from hardware appliances

In the previous section we talk about the move to IP and software workflows. But still most companies sell hardware appliances. We think this is going to change. After the contribution encoder, a dedicated link can bring your signal to the cloud.

There was a hype around cloud-based microservices but most people didn't seem to understand the concept. Most of the time you just don't need that level of complexity. We deploy microservices and know what we are talking about.