Norsk co-founder and CEO Adrian Roe and Streaming Learning Center‘s Jan Ozer explored the complexities of codec implementations, encoding efficiency, and the trade-offs between quality and cost in a recent webinar called “Cracking the Code(c): It’s All About the Implementation.” Here’s a breakdown of their insights and key takeaways for anyone navigating the world of video encoding workflows.
Why Codec Implementation Matters
The efficiency and performance of a codec depend heavily on how it’s implemented. For example, while AV1 is often touted as 50% more efficient than HEVC, whether you can achieve that efficiency depends on multiple factors.
Jan highlighted this variability, explaining that “with software, you can get 50% better efficiency with AV1 than you can with HEVC, but with hardware, the numbers are much closer.” This discrepancy is why many live streaming services haven’t fully adopted AV1 yet. “If the advantage of AV1 isn’t as great as what you think it is, the cost of switching becomes much higher,” Jan said.
Quality vs. Bandwidth: The Cost of Compromise
One of the most compelling parts of the webinar was the discussion on the relationship between quality and bandwidth costs. Jan emphasized that “the cost of quality is bandwidth.” In other words, if an encoder delivers lower quality, you’ll need to increase the bitrate to achieve the same visual fidelity, which drives up bandwidth costs over time.
To illustrate this, Jan shared a case study involving a 100-channel FAST service. He explained that “if you use a lower-quality encoder, you pay an additional bandwidth cost during the useful life of that encoder.” Jan shared results from a Moscow State University encoding comparison reporting that an NVIDIA encoder could deliver the same quality as an Intel Arc encoder at a 23% lower bandwidth. This translated to $170,000 in savings over an estimated five-year useful life of the encoders.
Adrian added that “consumers behave in a way where they consume up to a bandwidth far more than they consume up to a quality.” This means that while improving encoding efficiency can reduce network costs, it often leads to more high-quality streams being consumed, which can offset some of the savings.
The Role of Presets and Encoding Settings
Another critical topic was the impact of encoding presets and settings on quality and efficiency. Jan shared insights from his research on H.264 presets. One finding was that the veryfast preset saves you 94% of encoding costs as compared to veryslow, but you have to boost the bitrate of the veryfast encode by close to 200% to equal the quality you get with veryslow. Once you have delivered more than 50 – 100 hours of that video, the additional bandwidth costs exceed the encoding savings. This trade-off between encoding time and bandwidth costs is crucial for businesses to consider.
Adrian commented, “the devil is in the details” when it comes to tuning codecs. He explained that while presets provide a good starting point, “going down to the next layer of detail can be useful for optimizing performance.” He cautioned against over-complicating settings without testing, before citing an example where deep optimizations in codec settings helped increase compatibility from 16% of target devices to 96%.
Audience Size and Use Cases: One Size Doesn’t Fit All
Adrian and Jan also explored how audience size and use cases influence codec choices. Jan pointed out that “if you’re doing the Super Bowl, you’ve got 50 million viewers, and you want to make that stream as tiny as you possibly can because you’re paying to distribute it to every one of the viewers.” In contrast, smaller audiences should focus on delivering a premium experience rather than minimizing bandwidth.
Adrian agreed, noting that “if you’ve got large audiences, then if you can bring your average bitrate down, that’s likely to be the dominant factor.” He also highlighted the importance of the bottom and top rungs of the encoding ladder in particular. You can’t exceed the bitrate of the top rung, and the quality and bitrate of the bottom rung define the user experience in poor network areas—whether they can receive a stream at all and how it looks. One audience member agreed, pointing out that “serving decent-looking video to that bottom 2% of your audience on crappy connections is really important.”
The Future of Codecs: AI and Beyond
Looking ahead, both speakers touched on the future of codecs and encoding technologies. Jan speculated that AI-based codecs could revolutionize applications like autonomous driving, where traditional codecs optimized for human vision may not be suitable. He noted that “50% of all video is for machines these days,” suggesting that AI-driven solutions could offer significant efficiency gains for machine vision applications.
Adrian expressed concern about the industry’s focus on AI, warning that “the whole rush to ‘If it’s not AI, why the hell are you doing it’ is going to be bad for competition.” He argued that the media and entertainment industry needs more innovation in encoding technologies, but the high cost of developing specialized hardware makes it a challenging market.
Key Takeaways
Based on the webinar, here are some actionable recommendations for businesses navigating the codec landscape:
- For Large-Scale Streaming Services:
– Invest in high-efficiency codecs like AV1 or HEVC to reduce bandwidth costs.
– Use ASIC-based hardware for high-density, energy-efficient encoding. - For Smaller-Scale Operations:
– Focus on delivering a high-quality user experience rather than optimizing for bandwidth savings.
– Consider cloud-based solutions for flexibility and scalability. - For Developers:
– Benchmark codecs and encoders in the context of your specific use case (live vs. VOD, target devices, etc.).
– Use simple encoding presets unless testing confirms that advanced settings provide significant quality improvements. - For the Industry:
– Encourage competition in the encoding hardware market to drive innovation and reduce costs.
– Explore AI-based codecs for emerging applications like autonomous driving and machine vision.
Balance Technology and Business Considerations
The webinar underscored that there’s no single, perfect solution when it comes to codecs and encoding. As Jan aptly put it, “There is no one-size-fits-all codec comparison. There are just implementation comparisons.” (For an even more detailed discussion of codec comparisons vs. codec implementation comparisons, see this article from Jan.) Businesses must carefully evaluate their needs, audience size, and use cases to make informed decisions that balance quality, cost, and efficiency. That might mean testing multiple implementations of the same codec in your specific context rather than relying solely on vendor claims or third-party reviews.
In a world where bandwidth costs can make or break a streaming service, the right codec choice isn’t just about technology—it’s about strategy. As Adrian commented, “Choice of codec should be a business decision. Choice of codec implementation should be a business decision. Both make a big difference.”
