Video Quality Comparison for Various Codecs at Different Resolutions and Bitrates (Part 3)

In Part 1 and Part 2 of our video quality comparison series, we specifically focused on the H.264 video codec and how videos encoded with it perform under different conditions. While H.264 is still the most popular video codec used today, with 90% of the video industry supporting it, other codecs are increasing in popularity due to their efficiency, cost of implementation, and other factors. 

In this blog, we will expand the discussion by closely examining the successor of H.264, the H.265 (HEVC) codec, along with the open-source VP8 and VP9 codecs, which are commonly used in various web related video application scenarios.

H.264, H.265, VP8, VP9

For this experiment, we will be using a static video to investigate the performance of four different codecs: the most common video codec H.264, its higher-quality follow-up H.265, and VP8 and VP9—open-source, royalty-free standards currently owned by Google.

H.264

Advanced Video Coding, better known as H.264, is currently the most popular video compression standard in use.  First published in August 2004, it can encode high-quality videos at lower bitrates and is used in many streaming scenarios, both live and on-demand broadcasting. H.264 is compatible with almost all streaming protocols in use today and works with various video container formats, such as MP4, MOV, MKV and others.

It is, however, a patented compression standard that requires royalty payments for it to be implemented.

H.265

High Efficiency Video Coding (HEVC) standard, also known as H.265 or MPEG-H Part 2, is the successor to the widely used H.264 codec. It was first introduced in 2013 and aimed to provide better compression than its predecessor. HEVC is most commonly used in very high-quality streaming and broadcasting, as it can support resolutions up to 4K and even 8K, while achieving up to 50% smaller file sizes at the same resolutions as H.264.However, just like H.264, HEVC comes with licensing fees, and is generally more complex, requiring more processing power than older standards like H.264. This means it does not work on older devices without additional updates.

VP8

VP8 is a video coding format initially released as a proprietary standard in 2008 by On2 Technologies. On2 was later acquired by Google in 2010 and VP8 was made open-source and royalty-free. It shares many similarities with H.264 and is most commonly used in various browser-based applications, as it is the default codec for WebRTC traffic.

VP9

Shortly after releasing VP8, Google started work on the next iteration of the codec and in 2012 released VP9. Similar to H.265, VP9 can achieve up to a 50% improvement over H.264 and VP8, and is used by Google and Youtube for their 4K videos. 

Like VP8, it is most commonly found and supported by modern browsers. VP9 does, however, share some of the same drawbacks as H.265, requiring more processing and not being supported on older hardware.

Video complexity

Before we analyze the quality data, let's make sure that there are no significant differences between encoding with these codecs and compare the video complexity data. For these tests, we will be using a static video: a person sitting at a table with a fixed camera and in a fixed position. Just like we did in our previous comparison, we will use the Video Complexity Analyzer (VCA) to analyze the data.

As we can see, both the spatial and temporal complexity scores for all codecs are almost identical. We can conclude that our encoding settings should not have any baseline differences on the quality results.

Video and metric setup

Encoding

Comparison will be made between all four codecs: H.264, H.265, VP8 and VP9. All videos were encoded using FFMPEG in various resolutions and bitrates. We used the following resolutions in our comparison: 144p, 240p, 360p, 480p, 720p, and 1080p. And for bitrate variations we tested 200kbps, 500kbps, 1000kbps, 2500kbps, 5000kbps and 8000kbps, as they proved to be  breakpoints of data in most cases. For the codecs themselves, “libx264”, “libx265”, “libvpx” and “libvpx-vp9” FFMPEG video encoders were used.

Another important point to mention is the container formats. Videos encoded with H.264 and H.265 use the `MP4` container, while those encoded to VP8 and VP9 use the `WebM` container. This is because H.264 and H.265 are not supported by the `WebM` container and are standardized for `MP4`. While VP9 can be supported by `MP4`, implementations for VP8, however, are very rare and the default container for it is `WebM`, with the same being true for VP9.

VMAF

Just like in our previous comparisons, the main metric we will use for these comparisons is VMAF—a full-reference video quality assessment metric developed by Netflix. And with our custom alignment algorithm, we can align all of the videos with the help of the ArUco markers. For consistency between all tests, they are all resized to 720p before further cropping of the selected ROI (region of interest), as that is the resolution we use when evaluating the most common scenario when a video is played on a laptop screen.

VQTDL

For a no-reference comparison we once more are using VQTDL, or Video Quality Testing with Deep Learning, which is an image quality evaluation algorithm developed by TestDevLab that correlates well with subjective human perception and was developed specially for real-time streaming and WebRTC products.

Video quality results

Average results

Let's start by looking at the average VMAF results across all codecs:

Up to 360p, we can see a clear pattern. VP8 tends to produce lower scores at lower bitrates but catches up as bitrate increases, whereas H.264, H.265 and VP9 show similar performance through all bitrate variations.

Starting from 480p we can already see changes. VP8 can no longer keep up with the lower bitrates and the average score for 480p 200kbps is lower than it was for 360p. A small gap in scores also starts for H.264 at 200kbps.

If we take a look at VQTDL, a similar pattern appears, both at 360p, where VP8 trails behind the other codecs, and at 480p 200kbps, which has lower scores than 360p 200kbps.

At 1080p there is a much clearer difference between codecs. H.264 is not able to maintain quality at lower bitrates and there is almost a 10% difference at the 200kbps results compared to H.265 and VP9. VP8 performs the worst at this resolution, with VMAF scores at 200kbps bitrate more than 30% lower than the highest H.265 and VP9 results. However, VP8 is able to catch up with these codecs at 1000kbps.

When taking a closer look at each of the codecs, even more differences can be noticed. An interesting observation is how each codec has a different bitrate threshold for the highest 1080p resolution to equal and overtake other resolutions. At lower bitrates, there just isn't enough data for a 1080p video and so we can see that scores are the highest at 720p.

For H.264, this threshold is at 2500kb/s, where 720p and 1080p scores are almost the same.

H.265 performs better in this regard, with 1080p already achieving the best scores at only 500kb/s bitrate.

VP8 surprisingly performs better in this scenario than H.264, with 1080p being able to surpass other resolution quality scores at 1000kb/s bitrate.

And just like we have observed before, VP9 closely follows H.265 with the 1080p resolution also requiring only 500kb/s to equal lower resolutions.

Next, let’s look at some comparisons between the codecs themselves. For this analysis, we will be using H.264 as the reference codec, as it is still the most popular one of the four and compare the relative scores of the other codecs against it.

In this graph, we can observe the relative VMAF score difference between H.264 and H.265. If the scores are positive, it means that H.264 has scored better in that case, while negative scores show that H.265 outperformed H.264.

At lower resolutions, both codecs perform almost identically. This is as expected, because at these lower resolutions, there is only so much data that can be encoded in the video and even with more efficient codecs, the increase in bitrate yields very minor quality improvements.

We can see a small divide in scores starting from 360p and a more noticeable one at 720p 200kbps, as well as even larger ones for 1080p 200kbps and 500kbps, where H.265 starts to outperform H.264 more and more. However, at higher bitrates H.264 is able to catch up with H.265 quality.

Next, we look at VP8, where H.264 matches or outperforms it in almost every scenario. It is especially noticeable at low bitrates, where VP8 can score almost 25% lower in some of the higher resolution tests. As you can see in the example below, VP8 has quite visible blockiness and colour degradation, with slight spatial distortion as well, which contributes to the much lower quality scores.

When comparing H.264 and VP9, the relative scores are quite similar to the H.265 comparison. At lower resolutions, H.264 and VP9 have very similar scores throughout all of the bitrate scenarios. At higher resolutions and lower bitrates, H.264 performs slightly better than it did against H.265, but still lower than VP9.

Overtime results

Now let’s take a closer look at each codec's overtime results.

Starting from the lowest resolution at 144p, each codec produces relatively low but consistent scores across different bitrates:

The only anomaly in data comes from the VP8 144p 200kbps test, which has slightly lower scores than the rest of the tests for 144p VP8 tests, while the other codecs are able to maintain almost identical results for all bitrate variations.

If we investigate this specific case, we can see that there is a small but visible difference in comparison to, for example 500kbps, where you can see more detail around the top of the person's head and eyebrows, while in the 200kbps test that part of the face is more blurry.

As we increase resolutions, the lower bitrate videos start to fall behind in quality. However, as noted in previous blogs, more bitrate does not mean an increase in quality. At some point, just like we saw at 144p, there are diminishing results and a maximum quality that can be obtained.

At 480p, the difference between the lower bitrate videos becomes even more distinct, with videos at 200kbps lagging behind for all codecs, while older codecs like H.264 and VP8 also show a more distinct gap in results between 500kbps videos. Higher resolutions with 1000kbps tests also start to show signs of falling behind the higher codecs.

This gap in quality increases with each higher resolution, however, as mentioned before, we can also observe diminishing results from the higher bitrates.

At 1080p, each codec is able to achieve extremely close to maximum quality without much difficulty at 2500kb/s, meaning that trying to force 5000kb/s or 8000kb/s is wasted effort.

The newer H.265 and VP9 codecs are able to achieve near-maximum scores in this scenario with only 500kb/s.

Looking at the overall result patterns, VP8 clearly shows the most unstable data, stabilizing only at 2500kbps and higher bitrate for higher resolutions. In contrast, the other three codecs maintain stable data throughout almost all bitrate variations with H.265 being the most consistent overall. This pattern is present in both VQTDL and VMAF results.

Future of codecs

Before we get to the main conclusions, we must discuss what could come next for these codecs. While H.264 remains the most widely supported codec and VP8 and VP9 being prevalent in web-related spaces, a new contender has emerged in recent years—AV1.

AV1 is a royalty-free, open-source codec released in 2018 and developed by the Alliance for Open Media (AOMedia). It was designed to be more efficient than H.264 and VP9 by delivering high quality video while using significantly less bitrate. The adoption of AV1 has been steadily increasing over the years. Its main competitor is VP9, as both target high-resolution (4K or 8K) online streaming.

Some of the major advantages of AV1 are:

• Royalty- and licensing-free
• High quality even at very low bitrates
• Great compression efficiency (claims of 30-50% more efficient compression than H.264)

However, as a newer codec, there are some challenges and limitations that come with it:

• Requires more powerful hardware for encoding, limiting support on older devices
• Slower encoding that requires more processing power
• Limited hardware support

Even with these limitations, AV1 adoption has been steadily growing. It has been adopted by more and more mobile devices and other hardware. For example, in 2022, Intel Arc Alchemist GPUs were the first to feature AV1 encoding.Since then it has been featured in other NVIDIA and AMD GPUs. Major companies, such as Google, who are actually one of the major contributors to its development, and Netflix have been gradually adopting it into their workflows (at the end of 2025, Netflix reported that AV1 powers approximately 30% of all their viewing). 

Additionally, AOMedia has announced the release of AV2, the successor to AV1, with its launch being planned sometime in the first half of 2026.

While older codecs will still remain prevalent for their compatibility with older devices and lower processing requirements and implementation, organizations who want to future-proof their media workflows should look into adopting newer and more efficient codecs such as AV1, which has clearly shown improvements over the older generation of codecs.

Key takeaways

The data we collected for our video quality comparison provides us with valuable insights into each of the discussed codecs and how they perform under various bitrate and resolution scenarios, as well as a general understanding about each of the codecs. Here are some of our key takeaways:

Open vs. Licenced codecs

Patented codecs like H.264 and H.265 will often receive better early support and industry standardization, however, they can suffer from legal and licensing fees that can hamper their adoption if they are not thought through. Open-source codecs such as VP8 and VP9 can lag behind if there is not a major industry player supporting them, but they escape the troubles of licensing and legal constraints. While patented codecs, especially H.264, have historically been the most implemented solution, newer open and royalty-free codecs such as AV1 have shown that this approach might take over in the future.

VP8 falls behind other codecs

Based on our results, VP8 in almost all cases proved to have the worst quality results, both on average and also when looking at overtime data, where it often struggled to maintain result stability. However, in cases where ease of implementation is the main concern and factors such as compression efficiency are not the top priority, it is still a very usable solution.

H.265 provides the best compression efficiency

During our testing, H.265 was a clear winner among the tested codecs. It showed the best overall performance with the measured data being stable and able to provide high-quality scores even on high-resolution videos with low bitrate.

Diminishing returns in high resolution video quality above 2000kbps

As was observed, all codecs could easily achieve maximum video quality with bitrate set to 2500kbps, at which point there were very minimal gains, with some of the more efficient codecs such as H.265 and VP9 being able to use even less bitrate due to their improved compression efficiency.

Base the required bitrate from your resolution requirements

At low resolutions, using more bitrate than necessary only increases processing time and resource usage, without providing a substantial improvement in video quality.

Overall, our analysis highlights that choosing the right codec depends on the balance between quality, efficiency, and compatibility. While older codecs like H.264 remain widely supported, newer and more efficient options like H.265, VP9, and AV1 offer significant gains in compression and quality, making them the best choice for future-proofing high-resolution video workflows.