With the booming video industry in recent years, there has been a significant growth in media applications such as short videos, social media, e-commerce, and video conferencing. The demand for high-quality and low-latency content has become increasingly strong.
Tencent MPS focuses on two major functionalities: "Top Speed Codec" and "Video Quality Enhancement". It is a video and audio technology brand that combines Tencent Cloud's leading encoding, decoding, media processing, and AI technologies. In various media applications, Tencent MPS provides users with superior video quality and lower bitrate media processing services.
The overall processing pipeline of Tencent MPS differs significantly from conventional media processing workflows. After video decoding, we conduct pre-analysis processes such as scene analysis, artifact detection, noise detection, and interlacing detection to assess the visual quality of the video source. Based on the specific scene and visual quality conditions, we apply corresponding video enhancement/restoration techniques. After repairing the video source, Tencent MPS performs a secondary analysis of the visuals to assist the subsequent video encoding process. This analysis includes assessing the video's ROI/JND information and content-adaptive encoding information. Utilizing this information, we optimize the encoding process to be more in line with subjective human visual perception.
During the encoding process, we have also conducted in-depth optimizations of the encoding core. Through collaborative efforts across multiple departments within Tencent, we have developed proprietary encoders such as O264/V265/TXAV1/O266, which significantly improve video compression rates compared to open-source encoders.
First, let's introduce what Top Speed Codec means. The goal of Top Speed Codec is to minimize video bitrate while maintaining or even enhancing the visual perception of the human eye. This helps save bandwidth and storage. Top Speed Codec is achieved within the framework of Tencent MPS, which includes several steps such as video pre-analysis, preprocessing, and video adaptive encoding. Compared to regular transcoding, it can reduce bandwidth consumption by over 50% while improving the subjective visual experience to a certain extent.
The compression performance achieved by Top Speed Codec has undergone multiple iterations and optimization processes. In the initial development of Top Speed Codec, our focus was on utilizing the existing encoder and video processing capabilities effectively. We discovered that human perception varies across different scenes. For example, in certain games with extensive grassy areas, we can reduce the bitrate in texture-complex regions through parameter control in encoding. Although this may introduce ringing artifacts and aliasing in the texture area, the complex texture occlusion effect prevents the human eye from perceiving these artifacts. By analyzing the characteristics of the current scene and leveraging the features of video encoding, we can maintain the visual perception of the human eye at a lower bitrate.
As our business continued to iterate and develop, we realized that while open-source encoders have shown good results in the industry, they often struggle to fully meet the requirements of practical business scenarios. Additionally, open-source encoders may not implement all the features of the standard and may not achieve the maximum compression potential defined by the standard. Therefore, in the second phase of optimization for Top Speed Codec, we focused on algorithmic tuning of the encoding core. Our self-developed O264 encoder achieves over 20% coding gain compared to the open-source X264, while V265 achieves a coding gain of 40% compared to the open-source X265.
Top Speed Codec combines the power of AI to perform preprocessing on the video source before encoding, making it more suitable for the encoding scenario. The preprocessing is based on pretrained models that smooth out and eliminate edge details, resulting in smoother overall edges that are more conducive to video compression. Additionally, when the encoding bitrate is set to a low value, the model can estimate the compressed video, which helps mitigate block artifacts and excessive noise caused by insufficient bitrate. By simplifying complex textures and smoothing the video from a subjective perspective, the video becomes easier to compress and maintains good quality even at low bitrates.
The benefits of using Top Speed Codec are significant. Taking Tencent's internal business as an example, the use of Top Speed Codec has saved approximately 70% of storage and bandwidth costs. Additionally, due to the reduced file size, the initial loading time of videos has decreased by 20%, resulting in a greatly improved overall playback experience.
For extreme compression in on-demand scenarios, Top Speed Codec achieves impressive results. For example, with H.264, a 1080p HD movie video can maintain overall subjective clarity at a bitrate of 1.5 Mbps or achieve a VMAF score above 95. H.265 can achieve the same result at 900 kbps, and AV1 can even achieve it at 650 kbps. In educational scenarios, where there are often static frames, the compression effect is even more pronounced. For classroom scenes with PowerPoint presentations, H.264 can maintain subjective clarity at 67 kbps, H.265 at 35 kbps, and AV1 at 28 kbps. At this point, the video bitrate is mostly lower than the audio bitrate, significantly reducing storage and bandwidth requirements.
Another notable feature of Tencent MPS is its utilization of Tencent Cloud's massive global resources, enabling worldwide service deployment. Processing clusters are available in various regions to comply with local laws and regulations, supporting enterprises in their global expansion efforts.
For Tencent MPS, the future development of Top Speed Codec revolves around two main axes. Firstly, there is a deeper integration with AI capabilities. In the preprocessing stage, AI will be utilized to further enhance video quality to assist in compression. Within the encoder, AI capabilities will be leveraged to accelerate RDO analysis and prediction. On the playback side, efforts will be made to support LCEVC and implement super-resolution and quality enhancement through end-to-end testing.
Secondly, there will be a focus on optimizing for live streaming scenarios. Currently, many encoding tools have high complexity, making it challenging to ensure real-time performance in live streaming scenarios. Further optimization and acceleration of these encoding tools will be pursued to better serve the needs of live streaming applications.
With the acceleration of industry digital transformation, the era of seamless integration between online and offline, and the fusion of digital technology with the real world is rapidly approaching. Tencent MPS will share new industry trends, new technological directions, and new application scenarios in the era of seamless integration on our website, inviting everyone to explore the vision and create the future together!
You can experience the excellent effects of Top Speed Codec by Trying Demo. Welcome to Contact Us for consultation.