Workflow
Soul App全双工语音大模型升级 让AI在交互中拥有“主动性”
Zheng Quan Ri Bao Wang·2025-07-04 09:46

Core Insights - SoulApp has upgraded its end-to-end full-duplex voice call model, redefining the "full-duplex" interaction paradigm by eliminating traditional voice activity detection (VAD) mechanisms and delay control logic, allowing AI to autonomously manage conversation flow [1][2] - The upgraded model enhances AI's capabilities to break silences, interrupt users appropriately, and engage in parallel discussions, creating a more immersive and realistic voice interaction experience [1][3] - The company aims to leverage this technology for both one-on-one and multi-user interaction scenarios, enhancing the construction of diverse social networks [1][2] Company Strategy - SoulApp's CTO emphasizes the dual relationship of emotional and informational value exchange in social interactions, aiming to provide smarter and more immersive user experiences through innovative technology [2] - The company has been proactive in integrating AI into social networking since its launch in 2016, focusing on reshaping relationship discovery and expression in a pressure-free environment [2][4] - Soul has developed various AI capabilities, including language models and voice generation, and plans to further enhance its multi-modal end-to-end model by 2024 [2][3] Market Trends - The industry consensus anticipates an "AI application explosion" by 2025, with social networking being identified as a key area for AI deployment due to its high traffic value and network effects [2] - A recent report indicates that nearly 40% of young people use AI products daily for emotional companionship, with a significant increase in those willing to form friendships with AI, rising from 32.8% to 71.1% [3] - The demand for AI to exhibit more human-like interactions is growing, with about 60% of users expressing a desire for AI to feel more like real people [3]