Masterclass: Lessons in latency
Posted on Jul 26, 2025 by FEED Staff
In an era where real-time content is king, latency remains one of the biggest challenges in both broadcast and streaming
The panel
- Dr Ciro Noronha, CTO, Cobalt Digital
- Venugopal Iyengar, COO of digital, Planetcast
- Matthew Williams-Neale, Vice president of marketing and communications, Appear
- Dan Pisarski, CTO, LiveU
- Stefan Lederer, Co-founder and CEO, Bitmovin
- Evan Statton, VP of strategic business development, Zixi
What are the biggest bottlenecks in achieving true real-time streaming, and what tools are available to varying broadcast segments?
Ciro Noronha: There are two primary bottlenecks for low-latency streaming distribution. Encoding latency, both in the encoder itself and in the instructions it gives to the decoder for how much buffering it needs. This component can account for a few seconds of latency. In this case, there is a trade-off between latency, bit rate and quality. With the right equipment and encoding (HEVC ULL), this latency can be reduced without a quality impact if higher bit rates can be used.
The other is buffering latency through the path. This is added to ensure a smooth delivery to the end user. There is again a trade-off between network reliability and latency. Some of the newer protocols, such as LL-HLS, can help here, but the bottom line is that you can’t really reduce the latency unless the delivery pipes to the end users improve both in bit rate and quality. In other words, it is necessary to throw bandwidth at the problem.
Matthew Williams-Neale: Achieving true real-time streaming faces several bottlenecks, including network jitter, congestion, encoding delays and end-device variability. In live sports or interactive news, even milliseconds matter. The challenge is delivering high-quality content across diverse infrastructures while keeping latency imperceptible to audiences. To manage this, solutions must prioritise ultra-low-latency processing with highly efficient codecs such as HEVC and JPEG XS. Software platforms that support advanced protocols like SRT, enable seamless cloud integration and offer granular latency control from contribution to delivery are becoming critical for achieving ultra-low-latency streaming workflows. Ultimately, addressing latency requires a holistic approach across the entire workflow, tailored to each broadcasting need.
Dan Pisarski: A complete live production is not just the delivery of video out to the final viewer – the traditional role of a CDN and an area where technical innovations like LL-HLS have focused. It’s the complete video workflow from the lens at the event, through production and out to final delivery. LiveU started in the contribution part of that workflow, and here, latency is often a result of trade-offs related to resiliency. For example, before LiveU invented cellular bonding, satellite was a common contribution technology, but geosynchronous satellites carry a high latency requirement. Other forms of contribution, like dedicated fibre, may offer low latency but come at much higher price tags.
With cellular bonding, latency at the contribution stage has been pushed lower over the past decade as technology has improved. Today, combinations of technology such as 5G, LEO satellites and private cellular network (or private 5G) make the lowest latency contribution options available yet. The blend of these technologies is far from the high budgets of dedicated fibre or geo satellites.
Stefan Lederer: When it comes to real-time streaming, there are a number of bottlenecks that can affect latency, quality and encoding performance. When considering mission-critical content distribution, this is a serious issue, as content providers have a responsibility to ensure that sensitive content is distributed safely and efficiently. Among the most prevalent bottlenecks in real-time streaming is bandwidth limitations. For a content provider experiencing insufficient bandwidth, this can lead to dropped packets and increased latency, which is particularly problematic during high-resolution video streams. Another important consideration in real-time streaming is content delivery network (CDN) capacity. For a content provider dealing with an overloaded or poorly configured CDN, increased latency eventually decreases efficiency depending on server location. Bitmovin Player allows for content providers to stream closer to real time with low-latency Dash and HLS playback, creating a potential solution for the bottlenecks developing in real-time streaming.
Evan Statton: End-to-end streaming latency is a combination of encoding, transport, network and buffering. In order to achieve real-time streaming, all of these elements require tuning. Traditional codecs like H.264 introduce latency through multi-pass processing, while HTTP-based protocols (eg HLS, Dash) add buffering on both the sender and receiver. UDP-based protocols such as Zixi, SRT and Rist reduce transport latency through adaptive error recovery. Device and player buffers introduce delay unless carefully tuned. To manage these challenges, industries adopt tailored solutions. For example, sports and news customers may use the lowest- latency protocols possible such as JPEG-XS with Zixi, SRT or Rist for contribution and low-latency HLS with chunked transfers for delivery. Meanwhile, gaming and interactive platforms rely on WebRTC or custom software to enable bidirectional communication, which is even lower latency, but does not scale as effectively. Effective real-time streaming requires end-to-end coordination across the entire workflow.
"Solutions must prioritise ultra-low latency with highly efficient codecs such as HEVC"
How do emerging low-latency protocols compare, and what trade-offs should content providers consider when choosing a solution?
Venugopal Iyengar: Low-latency protocols come with unique advantages and trade-offs, all depending on the use case. LL-HLS (low-latency HTTP live streaming) is designed for compatibility with Apple devices and offers scalable delivery through traditional CDN infrastructure. However, its segment-based architecture can introduce small amounts of buffering, which may not suit highly interactive content. WebRTC, on the other hand, delivers sub-second latency, making it highly suitable for two-way interactions like cloud gaming, live auctions or betting. Its downside is a lack of scalability for large audience sizes due to peer-to-peer limitations and more complex server-side implementations.
SRT (secure reliable transport) is particularly strong in first-mile contribution, offering reliable, low-latency transport over unstable or variable networks. It’s ideal for getting content from the source to the cloud or production environments, but isn’t typically used for last-mile delivery to consumers. Many organisations opt for hybrid workflows, combining these protocols to align with specific delivery goals: using SRT for contribution, LL-HLS for mass OTT distribution and WebRTC for ultra-low-latency interaction. Choosing the right combination depends on latency tolerance, audience scale, interactivity and infrastructure compatibility.
Evan Statton: Each low-latency protocol serves a different need within the streaming landscape. LL-HLS enables delivery over HTTP with latencies as low as two or three seconds, making it ideal for large audiences with passive viewing requirements, such as live sports or events. It scales well with CDNs and adaptive-bit-rate workflows, but doesn’t support true sub-second interactivity.
WebRTC, by contrast, delivers real-time performance with sub-500ms latency and supports bidirectional communication, making it ideal for applications like cloud gaming, conferencing and live betting. However, it’s complex to scale because it is non-cacheable and therefore requires specialised CDNs to distribute effectively. Zixi, SRT and Rist offer secure, low-latency streaming over the public internet, with strong performance in challenging network conditions which include packet retransmissions, network bonding/multi-pathing and congestion management. The trade-offs typically involve balancing latency, scale, interactivity and operational complexity. Content providers must carefully assess whether the priority is broadcast-like reach, real-time responsiveness or contribution reliability and architect accordingly – often combining these technologies within a single workflow.
Dan Pisarski: Interest in reliable, low-latency delivery over different network types – such as public, private and wireless – has led to a growing list of protocol options. This is good news for workflows, but because some protocols were built for different parts of the content product chain (contribution, production, distribution) and due to protocols differing not just by technique but also by business and licensing models (as one example of the many differences between options), choosing the right protocol is not always straightforward.
The good news is that, increasingly, these protocols focus on low latency. Also, a positive in the world of protocols is that many originally designed for one part of the content chain have since been adapted for others – for example, the Whip implementation of WebRTC, which has transformed it from a delivery protocol into a viable option for contribution.
A trusted vendor can help eliminate some of the hard questions and offer a ready-to-use workflow for a large portion of the content chain – if not the full chain. This is partly because of adaptation of protocols as mentioned – for example, LiveU’s LRT (LiveU Reliable Transport) protocol, which excels at contribution, can be used for distribution through our LiveU Matrix cloud-native IP video distribution platform offering. Plus, vendors increasingly support a mix of protocols to fit the customer’s needs. At LiveU, this is part of our Ecosystem philosophy and has become increasingly embedded in our products.
"Low-latency protocols come with unique advantages and trade-offs, all depending on the use case"
Does traditional broadcast still outperform streaming in real-time delivery, and if so, what technological advancements could help close the gap?
Stefan Lederer: In today’s content landscape, traditional broadcast tends to outperform streaming in terms of real-time delivery, particularly for live events. This is primarily due to the robust, designed nature of traditional broadcast infrastructure, which can deliver content with minimal latency and high quality as originally designed. However, there are a few promising technological advancements that could potentially close
the gap for streaming in terms of latency.
The integration of 5G technology and networks could significantly enhance the quality and speed of live streaming. Low-latency protocols such as LL-HLS, WebRTC and SRT are designed to minimise latency, and adaptive-bit-rate (ABR) streaming allows for content providers to adjust the quality of the video stream in real time based on the viewer’s internet speed. Bitmovin Player and a variety of other Bitmovin streaming solutions allow for content providers to adapt their ABR streaming in order to fit specific needs, giving industry professionals a foot in the door ahead of traditional broadcast.
Ciro Noronha: Yes, traditional broadcast typically outperforms streaming because they have a dedicated air/cable/satellite channel. Closing the gap requires very low-latency contribution (using HEVC ULL), a suitable distribution protocol such as LL-HLS (supported end-to-end) and consumers that have appropriate network connection.
Matthew Williams-Neale: Traditional broadcast remains superior in real-time delivery due to dedicated infrastructure and deterministic transmission paths, which ensure minimal delay. Streaming introduces latency through various stages – buffering, adaptive-bit-rate selection and CDN propagation – particularly over public or heterogeneous networks. However, advancements are helping narrow this gap. Optimised encoding schemes, better edge compute integration and smarter transport protocols now enable sub-second streaming with robust reliability. Ensuring synchronisation, timing control and latency tuning across workflows is essential. Progress will continue as livestreaming infrastructures adopt more dynamic, modular architectures capable of handling high-throughput, low-latency video workflows – while offering the flexibility to adapt to changing demands.
"The method for syncing different video streams was not common"
How can broadcasters and streamers guarantee synchronisation across devices, networks and regions while maintaining low latency?
Dan Pisarski: The technique for synchronising different video streams – embedding timestamps in the out-of-band video data – was not common even a few years ago, but today it is (although not quite standardised). This technique involves adding global timestamps (such as those taken from NTP or PTP) into the headers of frames or GOPs for the video, so that different streams can have their ‘absolute time’ compared to re-align frames from production that occurred at the same time. Far more encoders support this technique today than ever before, and more production platforms use it to align frames – LiveU Studio’s cloud-native production solution is a good example. Sources of such timestamps are now easier to access, with PTP more common and NTP more accurate.
Venugopal Iyengar: Ensuring synchronised playback across devices and geographies while keeping latency low is one of the biggest challenges in live streaming. Frame-accurate synchronisation becomes difficult when viewers are using different platforms, network conditions vary and content is being processed and delivered through many nodes.
Solutions typically combine timecode watermarking, network-aware buffering and CDN-level delivery alignment. Time-aligned manifests and device-level playback controls can help reduce drift. Advanced scheduling platforms can manage stream variations and enforce timing precision.
Real-time analytics also play a critical role. By continuously monitoring latency variations and synchronisation issues, broadcasters can trigger corrective actions – adjusting delivery paths, changing buffer sizes or even switching between CDN nodes to maintain consistency.
Ultimately, combining AI-driven adaptive scheduling, automated monitoring and cloud-based orchestration gives a strong framework for maintaining sync while minimising delay, ensuring a seamless viewing experience.
"For experienced gamers, 100ms can be too long. For betting it is less critical, but the closer to the live edge the better"
With the rise of interactive streaming – such as live betting, cloud gaming and audience engagement – how critical is sub-second latency?
Evan Statton: Very critical, and sub-second isn’t low enough! Game streaming requires near real-time responsiveness in a bidirectional scenario. That is to say, a player must be able to press a button and have the resulting frame from that action be rendered, sent back to them and displayed all under 100ms. For experienced gamers, even 100ms could be too long. For betting, it is less critical to be within 100ms, but the closer to the live edge the better for the rightsholder because it means more bets can be placed.
In the betting example, sync is important as well, so players all have the same information. Audience engagement is the least sensitive of the three unless it includes video conferencing, which should be under 300ms.
Stefan Lederer: With the rise of new content formats, such as live betting, cloud gaming and audience engagement streaming (charity events, content creation) sub-second latency is becoming more critical by the day. For many applications hosting these new forms of content, the success and user experience of the content hinges on real-time interaction and responsiveness levels, particularly for engaging live streams.
Live betting relies on real-time odds and actions, and delays in the stream can lead to missed opportunities. It’s similar for interactive polls, competitive gaming and content creation formats. Specifically for competitive gaming set-ups, a delay can lead to a lag in the game, causing the player to fall behind, which can have detrimental effects in a competition. Charity streams rely on real-time audience engagement for donations, measures and Q&A features, which without sub-second latency, would not be possible.
Matthew Williams-Neale: Sub-second latency is key for the success of interactive streaming applications. In live betting scenarios, even slight delays can undermine fairness and integrity. Cloud gaming similarly demands low-latency performance to ensure responsive gameplay and a seamless user experience. Interactive broadcasts – where audiences vote, engage or switch camera angles – also depend on real-time responsiveness to maintain immersion.
Sub-second latency is no longer a premium requirement; it’s a baseline expectation for engagement-driven applications. For service providers, this requires a focus on latency optimisation – from transport layer to playback environment – ensuring that interactions are immediate, consistent and scalable without compromising quality or reliability.
Looking ahead, how will innovations like AI-driven compression, edge computing and 5G impact latency – and what challenges might they potentially produce?
Matthew Williams-Neale: AI-driven compression will optimise bit-rate decisions dynamically, reducing buffer requirements and improving quality under constrained conditions. Edge computing localises processing, compression, packaging and decision-making, cutting down end-to-end delay. 5G, particularly private networks, offers predictable, high-bandwidth wireless contribution, ideal for remote and pop-up productions. However, these innovations bring complexity: ensuring interoperability across AI models, managing orchestration at scale and maintaining consistent QoS on shared infrastructure. Broadcasters must implement effective and purpose-built solutions that offer end-to-end workflow visibility, adaptive resource allocation and robust security measures – supporting latency reduction without adding operational complexity.
Evan Statton: Of the three, edge computing likely has the greatest near-term impact on latency because it reduces round-trip time between the player and origin, pushing content and processing closer to the viewer. AI may bring gains, but many of the big wins – like encoder and buffer tuning – are already well addressed through existing methods. That said, AI could play a bigger role at the edge, making real-time decisions about routing, quality or protocol to minimise latency further. A major future opportunity lies in data replication and distribution models that bypass traditional CDN caching. Additionally, with increased bandwidth, I-frame only or JPEG encoding could eventually be used to further reduce latency at both ends.
Venugopal Iyengar: These technologies will be game changers. AI-driven compression reduces bit rates without degrading quality, cutting processing and delivery times. Edge computing brings content closer to the user, and 5G radically improves last-mile connectivity, enabling true mobility in low-latency streaming.
But these come with challenges: managing distributed edge nodes increases complexity, AI workflows demand robust training data and 5G integration raises questions around infrastructure interoperability. Our response has been to develop modular, API-first systems that allow plug-and-play with these emerging technologies, under the NexC umbrella. This ensures clients can evolve without being locked into rigid architectures.
Latency is no longer a fringe concern – it’s central to the future of streaming and broadcast. Through platforms like NexC, Planetcast is pioneering cloud-first, AI-augmented and protocol-flexible solutions that redefine the latency equation. Whether for IPL-scale events or interactive OTT formats, we’re enabling media companies to meet real-time demands with speed, scale and stability.
Ciro Noronha: None of these technologies directly help the latency; most of them may actually have detrimental effects. For compression, the more time you have to compress something, the better job you can do. Using AI-driven compression can perhaps increase your quality, but it is very likely increasing your latency. Really, the only thing that can help is high-quality bandwidth availability all the way to the edge. If you have that, you have a chance at reducing latency by playing with protocols. If you don’t have that, the battle is pretty much already lost. For this, the technology that can help is 5G, simply because it will give you more bandwidth.
This feature was first published in the Summer 2025 issue of FEED.
