The Progress of Video Codecs: From JPEG to AV1

Video compression has advanced faster in the last five years than in the previous two decades. If you are still thinking about video codecs in terms of H.264 versus H.265, you are significantly behind where the field actually is. This article traces the trajectory from JPEG-based approaches through to modern AV1, and then applies that lens to something concrete: how well do Russian video platforms actually perform?

The Codec Efficiency Curve

For a long time, the standard measure of codec progress was "twice as efficient as the previous generation every five to eight years." H.265 (HEVC) roughly doubled the compression of H.264 at equivalent quality. AV1, the open royalty-free codec developed by the Alliance for Open Media, again roughly doubled HEVC.

But the comparison that matters in 2024 is not between open standards — it is between open-source implementations and closed proprietary ones. Analysis of competitive codec evaluations shows that between 2021 and 2024, the top proprietary implementations achieved approximately 3x better compression than x265 (the leading open-source HEVC encoder) at equivalent perceptual quality. The best private AV1 implementation in current testing requires approximately 6.5x less bandwidth than legacy x264 at matching visual quality.

That is not a marginal gain. For a large video platform, the difference between x264 and a state-of-the-art AV1 encoder is the difference between building five CDN data centers and building one.

Who Is Winning the Codec Race

The companies at the top of codec efficiency rankings are almost entirely Chinese: Tencent, Alibaba, Huawei, and iQIYI dominate the leaderboards across multiple independent benchmark evaluations. This is not a coincidence — these companies stream enormous volumes of video domestically and have direct, measurable financial incentives to squeeze every bit of efficiency out of their encoding pipelines.

The gap between the best closed implementations and the best open-source tools is substantial. If you are encoding with open-source tools and your competitor is using a proprietary pipeline, you are likely spending roughly twice as much on bandwidth for equivalent quality. At scale, that is a company-ending cost disadvantage.

How Russian Video Platforms Compare

With that context established, let us look at the Russian market. The analysis tested YouTube, Vimeo, VK Video, and Rutube, plus 15 online cinema services including both Russian platforms and Netflix.

Rutube vs. YouTube

The headline numbers are uncomfortable reading if you work at Rutube. For unpopular content (videos that have not triggered the CDN caching that popular videos benefit from), Rutube requires approximately 95% more bitrate than YouTube for equivalent perceived quality. For popular content the gap narrows, but Rutube still needs roughly double YouTube's bandwidth.

Playback performance under poor network conditions is worse still: playback delay on Rutube averaged 8.5 times longer than YouTube in testing. This is partly an infrastructure issue and partly an encoding issue — Rutube's bitrate ladder has non-monotonic segments, meaning lower-quality renditions sometimes have higher bitrates than the rendition above them. This breaks adaptive streaming logic and causes freezes at exactly the moments when a user's connection is struggling and the player is trying to step down to a lower quality tier.

Bitrate variance on Rutube also occasionally exceeds 200 Mbps in bursts — a number that has no legitimate explanation for standard streaming video and points to encoding pipeline problems.

Online Cinemas: Netflix vs. the Field

The analysis of 15 streaming services tells a similar story at the premium end of the market. Netflix's AV1 implementation is the benchmark: exceptionally tight compression, minimal variance, consistent quality across the bitrate range. The worst-performing service in the test — a Russian platform — requires 14 times more bitrate than Netflix for identical perceptual quality.

That is not a 14% gap. That is 1,400%. For every 1 Mbps Netflix uses, that platform uses 14. A user on a 10 Mbps connection watching Netflix gets a better picture than a user on a 140 Mbps connection watching that service, all else being equal.

Most Russian streaming services are relying on outdated open-source encoding approaches: older x264 or x265 implementations without the custom perceptual optimization that major platforms invest in heavily. The services that have modernized their pipelines show significantly better results, but most have not.

Why Quality Measurement Gets Ignored

The author makes a pointed argument about why this situation persists. Standard MBA-influenced product management focuses on metrics like user retention, time-on-platform, and ARPU. Video quality is hard to measure and does not appear directly in any of those dashboards. The assumption is that "good enough" video is good enough.

This reasoning fails in two ways. First, quality problems are not invisible to users — they result in exactly the churn and reduced engagement that the retention metrics are supposed to track, just with a lag and no clear attribution. Second, the competitive threat from Chinese platforms — which have invested heavily in codec R&D and can offer dramatically better quality at lower cost — is very real. If those platforms expand into post-Soviet markets with optimized applications, the efficiency gap becomes a direct cost-of-business disadvantage for incumbents.

The Toyota Lean Manufacturing analogy the author uses is apt: Toyota's insight was that quality and efficiency are not opposites. Defects are expensive. Measurement precedes improvement. A platform that does not measure its compression quality cannot improve it, and cannot know when a competitor has lapped it.

What Good Looks Like

Netflix's AV1 pipeline represents the current state of the art for streaming. Its characteristics:

• Per-title and per-scene encoding optimization, not one-size-fits-all settings
• Minimal bitrate variance — tight control over peak-to-average ratio
• Aggressive use of the highest-efficiency AV1 encoding tools, including closed proprietary modules
• Continuous quality measurement against perceptual metrics (VMAF, SSIM) rather than just bitrate targets

None of this is secret. The Netflix tech blog has published extensively on their approach. The barrier is not knowledge — it is the organizational will to treat compression quality as a first-class engineering concern rather than an infrastructure cost to minimize.

For Russian platforms competing for domestic users, the gap is recoverable. For platforms that want to compete internationally — or defend against Chinese competitors who do — the window for catching up is not unlimited.