I wonder if you have ever experienced such a situation: you are happily "opening" with your teammates. When you were leading your teammates to kill all sides, your voice suddenly broke down. The original good situation suddenly fell into a crisis due to poor communication, which was simply collapsed!
As mobile games become increasingly popular, the sociality of games has become a necessity for players, and voice systems have become the top priority for measuring game quality, and game manufacturers also pay more attention to the development of in-game voice systems.
Recently, Tencent Game released the latest voice technology solution. The " Peace Elite " team of Tencent Photonics S studio, along with professional teams such as Tencent AI Lab and Tencent Game CROS GVoice (Tencent Game Voice), have been digging deeper into the upgrade technology of voice codecs, and applied AI Codec technology to the "Peace Elite" game, realizing AI Codec with lower code rates and higher quality voice encoding, which is far ahead in the industry and has become the first product to fully apply AI Codec technology to the field of game voice.
The Internet speed is different, and the poor voice "black" is a pain point
iMedia Consulting There was a statistics showing that 40% of players like a game due to social factors; 15.6% of players choose to withdraw because of insufficient sociality of the game. It can be seen that how to improve the sociality of games is the primary problem that major game manufacturers must face.
Effective communication in the game can increase the player's interactive experience, but with the development of mobile games, players have become increasingly demanding on the quality of the game's voice, delay time, and usage environment.
In fact, real-time voice of games has always been a cliché topic, and it is not just about real-time calls. When judging the quality of voice, the listening quality of voice is the top priority. This premise makes it easier for: the lower the transmission code rate, the less traffic the player consumes; the less computing power is used, the faster the CPU process.
Until today, the game voice transmission technology has been looking for breakthroughs in two aspects.
The first problem to be overcome is - how to ensure the stability of voice communication when the Internet speed is not smooth? Without reducing the quality of listening, the lower the voice information code rate, the more stable the transmission coherence. In layman's terms, low code rate has more advantages in weak networks. The loss of computing power of
is the second difficulty. Large games generally require high CPU requirements. Therefore, it is also particularly important to make the voice function occupy less CPU while not changing the player's communication experience, so as not to squeeze operations and resource calculations.
Artificial Intelligence -"BUFF" bonus for game voice
When voice black and game socialization gradually become a common demand for players, technology companies from various countries have also invested in this field.
Traditional audio decoding technology Codec focuses on reducing network bandwidth resources and transmission code rate to ensure that voice information is not distorted while ensuring voice quality. With the development of technology, traditional signal processing methods have gradually been replaced by voice encoders based on neural network , becoming the industry's leading technology. As a result, AI Codec, which evolved from artificial intelligence technology, has also attracted much attention from the industry.
The gaming industry has conducted many experiments in order to make artificial intelligence's "BUFF" blessing effect on game voice. For example, the Opus sound encoding format, which is well-known in the industry, and the ultra-low code rate AI Codec that appeared in 2019, but it is not satisfactory in terms of transmission and voice quality.
Google also involves this field. Although the AI Codec Lyra released in early 2021 can ensure that it does not distort under low-code rate transmission, the voice delay problem has never been solved. The AI Codec model SoundStream released in the middle of the year has improved compared to Lyra, but has poor compatibility and stability on most phones.
mobile AI Codec voice technology is launched for the first time "Peace Elite"
. At the same time, domestic manufacturers are not willing to be outdone and are committed to upgrading in-game voice interaction technology.
"Peace Elite" is a leader in domestic tactical competitive mobile games, and has always regarded improving game quality and player experience as its purpose. Due to the wide range and high frequency of game voice, the Peace Elite team is also very concerned about this.
The Peace Elite team jointly established a GVoice AI Codec R&D project with Tencent AI Lab and CROS gaming voice GVoice teams at the end of 2021. Based on existing technology accumulation, the new project pays more attention to the realization of in-game real-time voice environments with high quality, low code rate and low power consumption. Through targeted optimization of game scenarios, compared with traditional instant messaging software, GVoice AI Codec performs better sound quality in the in-game voice processing algorithm. Even in weak network environments, it can ensure lossless sound quality with low code rates, and is suitable for mobile devices.
It is reported that GVoice AI Codec realizes transmission bandwidth compression through self-developed codec model algorithms, and replaces the original 24kbps Opus with AI Codec at 5-12kbps, so that players' subjective auditory experience is not affected. This means that GVoice AI Codec can not only ensure sound quality, but also reduce traffic consumption. In terms of computing power consumption, GVoice AI Codec needs to reduce the complexity and resource consumption of operations through engineering measures such as model compression, quantization, parallel computing, and mobile instruction set adaptation. Data shows that testing GVoice AI Codec on OnePlus 3 duplex calls increased by about 2.3% compared with Opus format CPU occupancy resources; testing under iPhone 6s, the CPU occupancy rate increased by about 1.3% after using GVoice AI Codec. This shows that GVoice AI Codec uses extremely low CPU usage rate increase, achieving high-quality real-time calls with low traffic consumption.
(The picture on the left is the original voice, and the picture on the right is the voice coded after GVoice AI Codec. From the waveform, the voice fidelity of the right image is high, and the subjective listening feeling is close to the original voice)
GVoice AI Codec technology has been launched in "Peace Elite", and the quality of voice service has been greatly improved. Due to the optimization of voice traffic, GVoice AI Codec has greatly improved the voice fluency of players in weak network environments such as travel and outdoors, and the situations of kappa and severity in the battle situation are greatly avoided. In addition, GVoice AI Codec uses AI model algorithms, which will greatly improve vocal transmission compared to traditional voice algorithms. In "Peace Elite", players' voice clarity will also improve accordingly. Whether it is cooperation or chatting, it will double the social experience. After multiple rounds of comparisons in different noisy environments and multiple game scenarios, the project team received unanimous praise and feedback from players.
For the development of the game, GVoice AI Codec is applied to "Peace Elite", providing support for the research and development of more game voice technology, greatly increasing the possibility of related technologies being used in other games.
The game lacks social interaction is unsustainable. With the enrichment of game varieties, GVoice AI Codec will become a necessary technology for more games to enhance their experience and lead industry standards.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-