Tencent GVoice Technology
The Tencent GVoice is a game voice processing technology based on the fifth-generation Qualcomm AI engine, according to the official news from Tencent Games.
Team combat benefits come, and the voice game experience is improved. Noise reduction, better voice chat effect, and reduced performance consumption, Tencent game voice GVoice and Vivo use the fifth-generation Qualcomm AI engine to launch game voice processing technology to Enhance the game voice communication experience from the source.Qualcomm.
According to the official announcement, GVoice is optimized for gaming scenarios and has better voice quality than traditional IM software. The average noise duration of traditional IM software (VOIP) is 11.5 percent, while Tencent GVoice’s average noise duration is 3.4 percent. Running GVoice’s algorithm with the Qualcomm Hexagon DSP on the Qualcomm AI engine results in even better noise reduction. By moving the GVoicetopology position, Vivo deeply optimizes the GVoice parameters running inside the Hexagon processor on the Qualcomm AI engine to achieve better results for gaming scenarios.
With gaming and GVoice voice on, the noise reduction algorithm in GVoice is no longer running on the CPU, but on the Qualcomm AI-powered Hexagon processor, reducing power consumption by 20mA/s and reducing GVoice’s CPU resources by about 30%.
It covers real-time voice, voice messaging, text-to-speech, civilized voice, simultaneous voice transmission, magic voice change, voice announcements, and other application scenarios, and has been applied in such games as Honor of Kings, Peace Elite, League of Legends, Call of Duty, Tianya Ming Yue Knife, Perfect World, and so on, more than 2,000 internal and external games optimized.
Following are Official post from Tencent
Tencent’s GVoice, Vivo, and Qualcomm Snapdragon are the first to offer chip-level gaming voice processing.
To let players experience high-quality voice and smoothly open the game, Tencent GVoice and Vivo have jointly developed a game voice processing technology based on the fifth-generation Qualcomm AI engine.
In recent years, real-time interactive games are experiencing unprecedented rapid development, more and more players need to play as a team, and the demand for real-time voice transmission quality from players located in different locations is increasing. The smooth and rich gameplay voice communication has gradually become a necessary condition for the long-term operation of the game. Unlike traditional IM software, voice calls in hand games have higher requirements for “energy reduction and efficiency”.
(1) Users of traditional IM software will generally choose a quieter environment for their calls. The player is likely to be in a noisy environment when playing a game, so voice calls face the challenge of noise reduction from external noise, in-game background music, and sound effects.
(2) In general, the voice algorithm on a mobile phone will run on the CPU of the phone, which will consume a lot of CPU memory and other processing computing resources. The game itself already takes up a large number of computing resources, and it is necessary for the voice of the game to use minimal performance consumption to achieve the best voice experience.
To meet the huge amount of daily game interactions and reduce performance consumption while further improving the voice experience of players, Tencent Game Voice GVoice, together with global smartphone brands Vivo and Qualcomm, has embarked on a deep-seated technical optimization and launched a game voice processing technology using the fifth-generation Qualcomm AI engine to improve the voice experience from the source.
On November 2, Li Conobing, Deputy General Manager of Tencent’s R&D Department, and Yang Jinning, General Manager of Vivo’s Software Platform Department, held a signing ceremony in Shenzhen for the voice gaming technology cooperation. The signing marks the in-depth cooperation between the two parties in the field of game voice technology. At this stage, there are certain achievements in noise reduction and noise removal optimization, performance consumption optimization, etc. The two sides will continue to cooperate in echo cancellation, voice recognition, and other aspects.
Noise reduction and de-muting optimization
The GVoice algorithm is optimized for gaming scenarios, where it has better performance than traditional IM algorithms. While traditional IM software (VOIP) has an average noise duration of 11.5%, GVoice has an average noise duration of 3.4%. Running the GVoice algorithm with the Qualcomm® Hexagon DSP on the Qualcomm AI engine results in even better noise reduction.
By moving the position of the GVoicetopology, Vivo deeply optimizes the GVoice parameters running inside the Hexagon processor on the Qualcomm AI engine to achieve better results for gaming scenarios, as shown below for a certain scenario: compared to the traditional algorithm (left channel), Vivo and Tencent Games GVoice cooperation The speech chat custom optimization channel (right channel) has improved above the performance of voice reduction (cyan box) and noise suppression (yellow box), and its noise reduction effect can be improved by about 10dB in some scenes.
Sending voice effect optimization
When playing a game, the background sound effects are also a way to improve the gaming experience; similarly, voice interaction between teammates is very important; traditionally, the algorithm runs on the CPU, and to get a better background sound effect, the phone has to sacrifice some of the scenes to send voice effects, which sometimes leads to intermittent chatting between teammates; with GVoice, the phone and the game can be interrupted. The algorithm of the end is fully bound, Vivo can still get a better voice chat effect by deep optimizing the parameters of the voice algorithm running inside the Qualcomm AI engine Hexagon processor.
The left channel shows the voice chatting effect of the traditional algorithm, and the right channel shows the customized optimization channel of Tencent Game Chat between Vivo and Tencent GVoice), and for clear speech between teammates, there are no unintended effects (in the yellow box).
Performance consumption optimization
With gaming and GVoice enabled, the noise reduction algorithms in GVoice no longer run on the CPU, but on the Qualcomm AI-powered Hexagon processor, reducing power consumption by 20mA/s and reducing GVoice’s CPU resources by approximately 30%.
Based on the exploration and attempts in the field of game voice technology, this cooperation will accelerate the pace of commercialization of the core technology of game voice algorithms, empowering more and more game developers and ensuring that hundreds of millions of gamers can have a better gaming experience on the mobile platform.