Every byte you send to a viewer costs money — CDN egress, origin bandwidth, and storage for every rung of every ABR ladder. The codec you choose determines how many bytes it takes to deliver a given quality level. In 2026, the practical choice is between three codecs: H.264 (AVC), HEVC (H.265), and AV1. Each has a different cost-quality curve, a different device support story, and a different encoding cost profile.
This guide compares them from the perspective of teams that ship streaming apps to real devices — not from the perspective of codec researchers optimising for PSNR on test sequences.
The compression efficiency gap
At equivalent visual quality, the compression improvement over H.264 is roughly:
- HEVC: 30-40% bitrate reduction
- AV1: 40-50% bitrate reduction
These numbers come from production encoding benchmarks, not academic test sequences. Real-world results depend on content type, encoder tuning, and encoding speed. Sports content with fast motion sees smaller gains. Animation and talking-head content sees larger gains.
For a concrete example: a 1080p stream that requires 6 Mbps in H.264 to look acceptable might need only 3.5 Mbps in HEVC and 3 Mbps in AV1. At scale, that difference translates directly to CDN cost savings.
H.264: the universal baseline
H.264 (AVC) was standardised in 2003 and is supported by every device that plays video. Every smart TV, every mobile phone, every browser, every set-top box. If a device can stream video, it can decode H.264.
Strengths
- Universal hardware decode. Every target device has H.264 hardware decoders, including older Roku models, first-generation Samsung Tizen TVs, and budget Android devices.
- Fast encoding. H.264 encoders are mature and fast. Real-time encoding for live streams is straightforward and inexpensive.
- Proven ecosystem. Every packager, player, CDN, and DRM system supports H.264 without special configuration.
Weaknesses
- Compression efficiency. H.264 requires 30-50% more bitrate than newer codecs for equivalent quality. At scale, this is significant CDN and storage cost.
- No 8K support. H.264 was not designed for resolutions above 4K, and practical encoder performance degrades at very high resolutions.
When to keep H.264
Use H.264 as the lowest rungs of your ABR ladder to guarantee playback on every device. Use it as the primary codec when your device matrix includes hardware that cannot decode HEVC or AV1. Use it for live encoding when latency and encoder cost matter more than compression efficiency.
HEVC: the middle ground
HEVC (H.265) was standardised in 2013 and offers significant compression gains over H.264. It is well-supported on modern hardware but has a more complicated licensing landscape.
Strengths
- Good compression. 30-40% bitrate reduction over H.264 at equivalent quality. This is meaningful for 4K content where H.264 bitrates become impractical.
- Hardware decode on modern devices. Most smart TVs from 2018 onward have HEVC hardware decoders. All Apple devices since iPhone 7 and Apple TV 4K support HEVC. Modern Google TV devices decode HEVC natively.
- 4K and HDR. HEVC is the standard codec for 4K HDR content delivery via Dolby Vision and HDR10.
Weaknesses
- Licensing complexity. HEVC has multiple patent pools (MPEG-LA, HEVC Advance, Velos Media) plus individual patent holders outside any pool. The licensing situation has improved since 2020, but it remains more complex and potentially more expensive than H.264 or AV1.
- No browser decode without platform support. Chrome on desktop does not decode HEVC in software. HEVC in browsers requires hardware decoder support through the platform’s media pipeline. Safari supports HEVC because Apple controls both the browser and the hardware.
- Older smart TV gaps. Roku devices from 2016 and earlier do not decode HEVC. Some early Tizen TVs have limited HEVC profile support.
When to use HEVC
Use HEVC for 4K and HDR content where H.264 bitrates are too high. Use it as the middle tier of your ABR ladder for devices that support it, while keeping H.264 as a fallback. Use it when you need better compression than H.264 but your device matrix does not yet support AV1.
AV1: the efficiency leader
AV1 was developed by the Alliance for Open Media (AOMedia) and released in 2018. It is royalty-free, offers the best compression efficiency of the three, and has growing but still incomplete device support.
Strengths
- Best compression. 40-50% bitrate reduction over H.264, and 10-20% over HEVC. For bandwidth-constrained scenarios and cost-optimised delivery, AV1 is the most efficient choice.
- Royalty-free. No licensing fees. This is significant for large-scale services where HEVC licensing costs are material.
- Growing hardware support. AV1 hardware decoders are shipping in devices from 2022 onward: MediaTek and Qualcomm chips in smart TVs and mobile, Intel and AMD GPUs, Apple M3 and later, Samsung Tizen TVs from 2023 model year.
- Browser support. Chrome, Firefox, and Edge support AV1 decode (software and hardware where available). Safari added AV1 support in 2024.
Weaknesses
- Encoding cost. AV1 encoding is significantly slower than HEVC or H.264. Software AV1 encoding at production quality can be 10-50x slower than H.264, depending on the encoder and speed preset. Hardware AV1 encoders are emerging but not yet as mature or available as H.264 hardware encoders.
- Device support gaps. Older smart TVs (pre-2022) generally lack AV1 hardware decode. Roku devices before Roku OS 12 do not decode AV1. Many Samsung Tizen TVs before 2023 lack AV1 support. You cannot use AV1 as your only codec if your device matrix includes these models.
- Live encoding. Real-time AV1 encoding for live streams is possible with hardware encoders and fast presets of software encoders like SVT-AV1, but quality at live-speed presets is not as far ahead of HEVC as it is at offline presets.
When to use AV1
Use AV1 for VOD content where encoding time is less constrained and the CDN cost savings justify the encoding investment. Use it for browsers and modern devices that support it, as a high-efficiency tier in your ABR ladder. Consider it for mobile delivery where bandwidth savings directly improve the viewer experience on cellular networks.
Multi-codec ABR ladder design
In practice, most services in 2026 do not pick one codec. They run multiple codecs in the same ABR ladder and let the player select based on device capability.
A practical multi-codec ladder might look like:
| Codec | Resolutions | Target devices |
|---|---|---|
| H.264 | 360p, 540p, 720p, 1080p | All devices (universal fallback) |
| HEVC | 720p, 1080p, 4K | Modern smart TVs, Apple devices, Android |
| AV1 | 720p, 1080p, 4K | 2022+ smart TVs, modern browsers, mobile |
The player queries device codec support (via MediaSource.isTypeSupported() on web, or platform APIs on native) and selects the most efficient codec the device can decode. If the device supports AV1 hardware decode, it gets AV1. If it only supports HEVC, it gets HEVC. If it supports neither, it gets H.264.
This approach maximises quality per bit across the entire device matrix while ensuring every device gets a playable stream.
Encoding and storage cost
The tradeoff for multi-codec delivery is encoding and storage cost. Each additional codec multiplies your encoding pipeline. A 5-rung ABR ladder in three codecs means 15 renditions to encode, package, and store.
Offset this by:
- Encoding AV1 for only the top 2-3 quality rungs where the bitrate savings are largest in absolute terms
- Using per-title analysis to eliminate unnecessary rungs
- Sharing CMAF segments between HLS and DASH manifests to avoid duplicate storage
DRM and codec interaction
All three codecs work with CENC (Common Encryption) and CBCS encryption schemes. Widevine and FairPlay both support H.264, HEVC, and AV1 content.
However, DRM license acquisition and key rotation interact with codec selection at the player level. When the player switches codecs during an ABR adaptation (for example, from H.264 to HEVC because the player detected hardware HEVC support), the DRM session may need to be renegotiated depending on the platform’s EME implementation. Test DRM playback across codec switches on actual target devices, not just in Chrome DevTools.
For more on DRM testing methodology, see our guide on testing DRM playback across devices.
Encoding infrastructure in 2026
H.264 encoding is a solved problem. x264, Intel QSV, NVENC, and MediaCodec all produce production-quality H.264 in real time.
HEVC encoding is mature. x265, Intel QSV, NVENC, and Apple VideoToolbox all support HEVC at production quality. Real-time HEVC encoding for live is standard.
AV1 encoding is the frontier. SVT-AV1 is the fastest software encoder and is suitable for VOD at quality presets 4-6. For live, SVT-AV1 at preset 8-10 or hardware AV1 encoders (Intel Arc, NVIDIA Ada Lovelace) are viable but require careful quality validation.
The encoding cost per hour of content, including cloud compute or hardware amortization, looks roughly like:
| Codec | Relative encoding cost (VOD) | Relative encoding cost (live) |
|---|---|---|
| H.264 | 1x (baseline) | 1x (baseline) |
| HEVC | 2-3x | 1.5-2x |
| AV1 | 5-15x | 3-5x |
These costs decrease over time as AV1 encoders improve and hardware encoding becomes more available, but the gap is still significant in early 2026.
Practical recommendation
For OTT services shipping to a broad device matrix in 2026:
- Keep H.264 as the universal fallback. Do not drop it until you have verified that every device in your support matrix can decode a newer codec.
- Use HEVC for 4K/HDR content and as the primary codec for modern smart TVs and Apple devices.
- Add AV1 for VOD delivery to devices that support it, starting with the highest-bitrate rungs where the CDN cost savings are largest.
- Monitor codec selection in your analytics to understand what percentage of viewers are actually getting AV1 vs HEVC vs H.264, and use that data to guide future encoding investments.
The codec landscape will continue to shift as AV1 hardware decode becomes universal, but in 2026, multi-codec delivery is still the pragmatic approach for services that need to reach every screen.