iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
In one recent five-second 720p test, AV1 made the smallest file—but the codecs were not tested at equal visual quality. That result makes AV1 the smallest in that particular run, not a guaranteed winner for every video. The answer depends on the source, encoder implementation, settings, and quality target.
What the measured comparison found
Rendobar measured one encode of the same five-second, 1280×720 clip for each format, with audio removed. The runs used its API on August 20, 2026. The listed file sizes were:
| Format and encoder | Settings used | Output size |
|---|---|---|
| H.264, libx264 | CRF 23 | 433.0 KB |
| HEVC, libx265 | CRF 28 | 233.0 KB |
| VP9, libvpx-vp9 | CRF 32 | 346.6 KB |
| AV1, libsvtav1 | CRF 35, preset 8 | 224.4 KB |
| AV1, libaom-av1 | CRF 35, cpu-used 8 | 146.5 KB |
In this run, libaom-av1 produced the smallest output: 146.5 KB. But each encoder used its own CRF setting, and those settings do not guarantee equal quality. The size ranking therefore does not establish which codec would make the smallest file at the same perceived quality. The two AV1 encoders also produced noticeably different sizes despite using the same nominal CRF, illustrating why a codec label alone does not determine the result. Rendobar’s measured comparison and its methodology notes describe the test and its caveats.
Free tools Windows power users keep installed
One-click scans. No signup required.
The VP9 run did not use row multithreading or faster CPU-used/deadline settings; the author says it was not rerun with those flags. Timing was based on a single sample and can vary, so this test is not a dependable way to rank encoding speed either.
#1 Best Overall
Why “smallest” needs a quality target
A smaller file is only a useful win if its quality is acceptable. Constant-rate-factor (CRF) values are encoder-specific: the same number across different codecs—or even different encoders for one codec—does not mean the same visual quality. Meta makes the same point about quantization parameters in its real-time communications work: “QP is not comparable between AV1 and H.264 codecs, and hence can’t be used.” That statement concerns Meta’s RTC implementation, but the comparison principle applies here: a matching control value is not a quality match. Meta’s 2024 RTC account explains its context.
A useful codec comparison should encode identical source material at matched visual quality, then compare the resulting sizes. A second useful test fixes the output size or bitrate and measures which encode looks better. Rendobar also describes a separate 150 KB comparison assessed with VMAF, but it is a different experiment on limited material; it should not be read as a broad ranking across content. The comparison page discusses both tests.
Rank #2
- For size at comparable quality: match quality across codecs, then measure output size.
- For quality at a fixed delivery limit: encode to the same bitrate or size and compare visual quality.
- For a practical decision: also measure encode time and resource use, and check playback support on the target devices.
What broader comparisons do—and do not—show
Meta’s 2018 Facebook video study
Meta Engineering described encoding 400 popular public Facebook videos using a snapshot of the AOM AV1 reference software, FFmpeg 3.3.3 with libx264, and libvpx-vp9. Its method swept CRF/QP settings and then used two-pass average-bitrate encodes to compare quality and rate ranges. Meta concluded that AV1 could provide better quality at a given bitrate or reduce buffering at comparable quality in its service context. This is useful evidence that matched quality and bitrate matter, but the encoders are historical and the result belongs to Meta’s content and test conditions—not a current neutral consumer benchmark. Meta Engineering’s 2018 methodology and results.
Recommended Free Tools
Meta’s 2025 compression claim
Meta’s 2025 AV1 white paper says AV1 offers “up to 30% better compression than VP9.” That is Meta’s vendor-published upper-bound claim, not a result from the single-clip table above. The paper also notes that the efficiency gain comes with greater computational complexity and that playback can be challenging on low- and mid-tier phones with limited compute resources. Meta’s 2025 AV1 white paper.
How encoding cost and compatibility affect the choice
AV1 may reduce storage or delivery bitrate when an encode meets the quality target, but producing and playing that encode can cost more computation. In its 2024 mobile real-time communications implementation, Meta reported that AV1 used more memory and was disabled on low-memory devices. Its hybrid encoder could switch between AV1 and H.264 according to CPU use, battery level, or encoding time; Meta also reported a 2 dB improvement with AV1 using its locally developed PSNR framework. Those findings describe Meta’s RTC system, not a general file-encoding benchmark. Meta’s RTC engineering account.
Playback support also varies by device, operating-system version, profile, and hardware acceleration. Android’s platform documentation lists VP9 decoding from Android 4.4+ and AV1 decoding from Android 10+, and says AV1 encoder and decoder support are mandatory beginning with Android 14. These are Android platform statements, not a guarantee about every device’s performance or support on other operating systems. Android’s media-format support table.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Choosing settings for a real encode
For on-demand VP9 video, Google recommends constrained-quality mode with bitrate limits around a target. Its examples include 1,024 kbps for 720p at 24–30 fps and 1,800 kbps for 1080p at 24–30 fps. These are recommendations for Google’s web and mobile VOD use case, not universal quality guarantees or a direct comparison against H.264 or AV1. Google also notes that tiling can speed encoding, with a slight quality reduction. Google’s VP9 VOD encoding guidance.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →For your own content, choose representative scenes—not just a short, easy-to-compress clip. Include motion, texture, gradients, and the kinds of detail viewers care about. Keep resolution, frame rate, source, and quality target consistent; document the encoder and version, preset, and settings. Compare both the resulting quality and size, and test playback on the devices that matter. Without those controls, a smaller output may simply reflect a lower-quality encode or a more favorable test clip.
Quick Recap
So, which codec should you use?
- If you only need the result from the measured clip: libaom-av1 was smallest at 146.5 KB under its listed settings, but equal quality was not established.
- If you need the smallest file at a defined quality: there is no universal winner established by the available comparisons. Test the encoders and settings on representative content at matched quality.
- If playback reach or device cost matters: weigh device support, decode capability, memory, battery, and encode resources alongside file size.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
