Is Amazon bringing interactive live streaming to the masses? A closer look at Amazon Interactive Video Service (AWS IVS)
Correction: Twitch reached out to us to clarify that IVS runs on Twitch infrastructure, not AWS servers. I’ve updated the article to clarify this.
Amazon finally announced its long-awaited interactive live streaming solution, Amazon Interactive Video Service (AWS IVS). The solution promises to “handle everything from video ingesting and transcoding to global distribution for playback, so you can focus on building your own interactive application and audience experience”. The service claims to combine ultra-low-latency live video with interactivity, based on the same technology as Twitch.
Ever since Amazon acquired Twitch in 2014 for $970 million the companies have become more interwoven with several Twitch employees moving to Amazon HQ to work on video products. In 2017 Amazon started hiring for ‘Amazon Live’ to build interactive streaming solutions that create new shopping experiences for customers. In 2015, a year after the acquisition of Twitch, Amazon acquired Elemental for an estimated $350 million to strengthen its video encoding and distribution capabilities, making it part of AWS Media Services. Amazon claims that the new platform is a separate service from Elemental since it is purely focussed on web experiences, but all these steps have led up to the launch of Amazon IVS (AWS IVS).
However, interactive live streaming is nothing new. For decades people have tried to crack the code for the best live video experience by adding viewer engagement and data overlays for deeper immersions and a personalized experience. Both traditional broadcasters and online platforms have innovated how video content is consumed and how live experiences can be enriched — with varying degrees of success. Especially for the generation who now grows up with YouTube, HQ Trivia and TikTok these innovations are now the default viewing experience.
This post looks at the innovations and requirements for interactive live streaming and what Amazon’s IVS brings to the table.
1. Low latency streaming
Low latency live streaming reduces the latency between the source of the video and the viewer. With traditional live streaming solutions this latency can range anywhere from 15–120 seconds depending on the standards and technologies used. In recent years there have been significant developments to reduce the latency of existing and new streaming standards, opening up new opportunities for interactivity.
Low latency streaming means viewers are closer to the real event, making it more immersive. It can also prevent your neighbor from cheering before you even saw the goal. Low latency is especially key to interactive live streaming and viewer engagement as it allows viewers to respond to the host or events on stream and the feedback can be used directly in the content. Interactions can include poll and trivia questions, bets and predictions or chat messages.
Longer latency creates delays before a question reaches the viewers and adds delay before all the answers reach the studio. A feedback loop with high latency can make interactive live experiences feel slow and disconnected. For example, a live trivia game show with 20 second latency can double the time it takes to host trivia questions. The presenter asks the question, but it takes 20 seconds to reach the viewer, and then it takes a further 20 seconds for the answers to come in, before the presenter can respond. A lower latency creates a better viewer experience.
Another challenge with low latency streaming is that if a user has network connection hiccups this can result in the stuttering and buffering of the video stream. By lowering the latency between the source and the client the player has a smaller buffer to catch up when the connection drops. As a result the quality of experience and quality of service is more difficult to control. The example below shows viewing experience for different target latencies with an identical network situation. Finding the right balance between latency, video quality and user experience is key when providing interactive video.
Offering ultra low latency at scale and to a global audience requires a delicate balance between all these factors. Ultra low latency might work great in well connected regions like the US but can create hiccups for viewers in South America. Part of this is related to the CDN infrastructure delivering the video stream. Ultra low latency streaming requires a distributed network that provides multiple ingest points, pop locations and edge servers to reduce the distance and delay between the source and viewer.
Amazon IVS offering
Amazon IVS is based on a similar streaming experience as Twitch and can reach a latency as low as 3 seconds when using the Amazon IVS player on Desktop and Android. Third-party HLS video players are not supported and the IVS does not support low latency on iOS Safari browsers. You have to configure your own software or hardware encoder to achieve the lowest latency. Amazon IVS is based on existing Twitch infrastructure (a previous version of this article mentioned it was based on AWS infrastructure which has different availability in different regions which can have impact on the performance and scalability in less connected regions). However, because Amazon IVS does not actively sync the video between devices the observed latency can vary between users. For example, playing low latency video on iOS Safari can have 7–15 seconds of latency compared to 3 seconds latency on a desktop with a Chrome browser.
This brings us to the next requirement for interactive live video; a synced experience between viewers.
2. Synced experience
A synced experience ensures that every viewer sees the same frame at the same time. When you stream has a variable latency across devices, different viewers may experience different latency. Especially for longer streams this latency can start to drift between viewers. When adding interactivity to your livestream this latency can affect the viewer experience. For example in the trivia example when a question appears on top of the stream but is not mentioned by the presenter yet. It can also result in spoilers when data overlays are updated before the stream shows the event which can ruin the game when watching exciting sports and esports matches. This becomes even more important when there is money involved, for example with sports bettings, major financial announcements or live auctions.
To provide the best viewer experience for interactive live streaming you need to sync the video latency between viewers. When every viewer sees the same frame at the same time the interactivity or data overlays will always be displayed at the right moment. This way interactions can be provided and triggered from a third party platform without the need to integrate metadata in the video stream. If the video is equal for each viewer the interaction can come from any data or gamification platform. This makes the integration of interactive overlays more flexible and less dependent on a data stream in the video. If the video stalls the interactions are still displayed for the viewer.
If video is not synchronized the interaction and chat can be synced to timecode and metadata in the video stream. In this case the interaction or data is displayed when the interactive overlay receives a trigger from the video stream. Although this ensures that an interactive element is displayed on the correct moment in the video, the experience is different for each viewer depending on their latency. Interactions that are not in sync allow players to cheat and collecting results can take additional time.
Amazon IVS offering
Amazon IVS provides an HLS-based streaming service that can achieve low latency, but which is not synchronized between players. This is similar to Twitch in which latency depends on the client’s connection and settings. For example on Twitch the viewers can enable or disable low latency streaming depending on whether the streamer provides this option. As a result, latency between viewers can differ significantly. Twitch solves this by syncing the chat to the timecode but this is not an option for more advanced interactive live stream experiences or when real money is involved.
Additionally, there are limitations with low latency streaming on iOS web. All viewers that use their iPhone to watch a web stream will have additional latency depending on the stream configuration. Although Apple announced during WWDC that they will support some form of low latency HLS streaming this is not yet generally implemented. Amazon does not mention any support for low latency on iOS web.
So although Amazon IVS provides low latency streaming, the viewer latency is not synchronized between devices. This results in drifting latency values that can range between 2–10 seconds making true interactivity much more difficult.
Video interactivity is most effective when the input from viewers is incorporated into the format. This can be the results from a poll, the results from a trivia game, or the winner of a competition or auction. So let’s look at the last requirement for successful interactive live video; the interactivity.
Interactivity
Amazon introduced AWS IVS as the perfect solution for creating interactive video experiences. But what is an interactive video experience? At Ex Machina we have been developing interactive experiences for live events for many years, providing solutions for TV formats, Esport competitions and live stream projects. Our solutions range from trivia games, voting for talent shows, live esports predictions, and data overlays for sports. The key lesson from hundreds of projects is that feedback loops — acknowledging viewer feedback on screen — drives engagement.
Combining an interactive overlay on top of low latency live video makes the interactive experience more integrated and obvious to use. Especially compared to second screen apps or adding polls outside of the player (as we see in many conference tools). Creating a cross-platform experience with an interactive overlay on top of a live video can be challenging. A great example of integrated interactive live streaming are Twitch Extensions.
Managing and producing a live interactive experience requires a special set of tools to create, trigger and analyze interactions. These tools can get complex, especially when dealing with gamified experiences at scale, combined with points and leaderboards or when data is integrated from 3rd party APIs. Creating tools with an easy to use interface allows the production team to focus on the content.
Developing a custom interaction platform can be costly and challenging, especially at scale. You need to make sure interactions are triggered for each viewer at the same time, data is displayed at the correct moment of the video, and results are not visible too early to prevent cheating and spoilers. Most of the tools that Ex Machina has developed over the years run on AWS servers but are based on custom-developed server logic. This ensures that the entire chain is redundant, secure, and scalable for millions of viewers.
Amazon IVS offering
When we look at the interactive tools that Amazon is offering with IVS, the main feature is the ability to send custom timed metadata with the live stream to the player. This can be used to highlight information on the stream as they do with Xray on Amazon Prime. It can also be the trigger to display a poll, trivia question or prediction but IVS does not provide any of these capabilities by default. It seems that the interactive capabilities are very limited and are based on the PutMetadata API that “inserts metadata into an RTMPS stream for the specified channel”.
- The IVS PutMetadata API can only be added to RTMPS streams which limits the versatility of the platform.
- The data stream is only for publishing data and cannot be used for receiving results (e.g. answer to a poll). As discussed a feedback loop is the main driver for engagement.
- The PutMetadata API requires a third party solution to actually manage the content to display on stream. It’s up to the developer to create a standalone interaction platform. From stats and analytics to polls, trivia and predictions, additional development is required when using Amazon IVS of interactive live streaming.
The verdict
It seems that Amazon’s IVS solution provides tools for setting up a low latency live stream but is limited in supporting interactivity. This makes sense, Amazon Web Services is a build-it-yourself platform offering different components and does not provide an end-to-end solution. Anyone considering using Amazon IVS for interactive live streaming will need to work with partners or develop their interactive solutions in-house.
Amazon IVS is a great addition to the AWS suite to create a low latency live stream but seems limited when it comes to supporting true interactivity.
Ex Machina has developed Livery as an integrated solution that combines affordable and scalable low latency live streaming that is in perfect sync with the interactive tools that Ex Machina has provided for years. Ex Machina has years of experience in making live streaming interactive and was launch partner of Twitch Extension. For our projects we provide interactive overlays with real-time data, polling, trivia, betting, predictions and viewer engagement for sports, esports, entertainment and events.
Based on our experience we know that interactive live streaming requires low latency streaming, a synchronized video experience and two-way interactivity. Amazon IVS seems a great start to kick off a live streaming project but it is missing some key features to deliver true interactive, live video experiences, let alone to create the next Twitch platform. It does show that Amazon is committed to adding advanced video streaming tools to their portfolio. It will be interesting to see what their roadmap will look like.
Considering your own interactive live streaming project? Contact us to discuss how we can help you!
