GF SEC: BeanPod releases real-time voice model, optimistic about Byte industry chain + companion app user experience improvement.

date
21/01/2025
avatar
GMT Eight
GF SEC released a research report stating that the launch of the Dou Bao real-time voice large model further enriches the Byte AI ecosystem, optimistic about Byte AI's leading position in China and its future development potential. At the same time, emotional companionship and AI toy products are expected to enhance user interaction experience along with the strengthening of voice models. Event: On January 20th, the Dou Bao real-time voice large model was officially launched and fully opened on the Dou Bao APP. The model achieves integrated voice understanding and generation, end-to-end voice conversation, mainly targeting Chinese contexts and scenarios. Relying on voice and semantic joint modeling. According to the Dou Bao large model team official account, through external crowdsourcing, the model overall satisfaction is significantly superior to GPT-4o, with higher naturalness and emotional richness in voice tone than GPT-4o. GF SEC's main points are as follows: Pre-training Scaling + post-training high-quality data and algorithms bring breakthroughs in humanization, intelligence, expressiveness, and low latency in speech models. Traditional speech dialogue task systems have limitations in dialogue naturalness, usefulness, and security. According to the Dou Bao large model team official account, the Dou Bao team deeply trained various modal intertwined data during the pre-training phase, accurately captured and efficiently compressed massive voice information, achieved deep integration of voice and text capabilities to the maximum extent through Scaling, and emerged capabilities, in the post-training phase, using high-quality data and RL algorithms, providing the model with high emotional intelligence conversation capabilities and security. The Dou Bao real-time voice large model achieves personalized emotional connection, powerful voice control, rich emotional interpretation capabilities, a balance of intelligence and expressiveness, smooth interaction experience, and ultra-low latency. Leading large models and application layouts, focusing on opportunities in the Byte AI industry chain Currently, Byte's general large models and multimodal capabilities are leading among domestic major companies; according to AI Product Rankings, in December 2024, Byte's Apps Dou Bao and Cat Box reached 71.16/6.88 million monthly active users, becoming the top 1 in the monthly active users' track in China for AI dialogue and AI companionship. Byte continues to increase its investment in AI, and it is expected to have more technological and application advancements in the future. Companies related to Byte AI industry chain are expected to benefit. Enhanced emotional capabilities drive companionship experience upgrades, focusing on the effects of audio platform integration It is expected that emotional companionship products will improve along with the strengthening of model capabilities, with platforms with specific application scenarios expected to benefit. In particular, Dou Bao's conversation companion product Cat Box itself has the potential for an improved experience; for external platforms, platforms primarily focused on audio and dialogue are expected to enhance their service capabilities by integrating Dou Bao's emotional model; in addition, AI toys, as AI hardware with both educational and companionship attributes, are expected to significantly benefit from the empowerment of voice model capabilities. Investment recommendations: (1) Companies that have a cooperation foundation with Byte in the field of AI, such as Easy Click Worldwide Network Technology (301171.SZ), BlueFocus Intelligent Communications Group (300058.SZ), etc.; (2) Companies that are involved in AI voice-related businesses, such as Hubei Century Network Technology Inc. (300494.SZ) (music-social platform "Voice"), Kunlun Tech (300418.SZ) (AI music products Melodio and Mureka), etc.; companies that are involved in AI toys, such as Shanghai Film (601595.SH), Alpha Group (002292.SZ), etc.; (3) From an industry perspective, with the improvement of AI large model capabilities and the marginal cost reduction, continuously improving AI infrastructure is expected to better empower scenarios such as AI games, AI movies, AI marketing, emotional companionship AI education, AI agents, etc. Risk Warning: Model iteration effects may be lower than expected, commercial application landing may be lower than expected, risks related to copyright, ethics, and content quality, etc.

Contact: contact@gmteight.com