The toy industry has entered its “smart moment.”
According to data from the Ministry of Industry and Information Technology, the scale of China’s AI toy market reached approximately 24.6 billion yuan in 2024, and is expected to grow to 29 billion yuan in 2025.
As Sir Nigel Shadbolt, a professor of computer science at Oxford University, insightfully noted: “We are entering an era where artificial intelligence cares for the elderly and accompanies lonely children.” This means that future toys may not only be companions but also intelligent agents capable of understanding emotions and providing companionship.
From November 21 to November 23, the 14th Global Software Case Study Summit (referred to as: TOP100Summit), hosted by msup, was held in Beijing, and NetEase Cloud Communication’s emotional companion intelligent agent was selected as an annual case. In the session on “Empowering Product Innovation with Large Models,” NetEase Cloud Communication’s Product Director Que Hangning was invited to attend, sharing on the theme of “Innovation ‘Acceleration’: How an AI Capability Foundation Shortens the Industry POC Cycle from 4 Months to 1 Month.”
Based on a decade of technical accumulation in the instant messaging field and five years of in-depth practice in C-end children’s products, Que Hangning systematically explained how to achieve the integration of communication technology and intelligent interaction under the AI capability foundation, constructing emotional companion vertical scene solutions to help partners significantly reduce the innovation threshold and POC cycle of AI applications.
So, as a ToB enterprise that has a certain distance from end users, how does NetEase Cloud Communication choose its AI innovation direction? What choices and challenges did it face in the process? What practices and experiences does it have? Let’s review this wonderful sharing together..
01
Choosing the Innovation Track: “Communication + Intelligence”
Over the past decade, NetEase Cloud Communication has always resonated with the development pulse of China’s internet industry. When the “Internet +” strategy was proposed in 2015, NetEase Cloud Communication keenly chose to delve into the integrated communication track, solving the connection problems of “people to platform, people to people” for over 2.35 million enterprise customers. By 2025, under the national “Artificial Intelligence +” strategy, NetEase Cloud Communication believes that the integration of “communication + intelligence” will become a new infrastructure.
On one hand, from intelligent Q&A assistants to emotional companion products, from entertainment social interaction to this year’s highly anticipated AI toys, the essence of these innovative applications is to build a more natural and smoother human-computer interaction experience. On the other hand, the content of intelligent replies is expected to have more “human emotions”, as seen in AI toys like Ropet, which can understand emotions and achieve personalized interaction, are becoming the new favorites in the market.
The Ministry of Industry and Information Technology also stated at the “Introduction to the Revision of the Mandatory National Standards for Toy Safety Press Conference” that it aims to “promote the high-quality development of the AI toy industry,” which also proves that AI toys have been officially included in the national industrial narrative.
02
Ongoing Challenges Faced by AI Toys
However, transforming AI technology into a smoothly functioning toy is not an easy task.
In the realm of physical manufacturing involving appearance, materials, structure, and molds, traditional manufacturers still have advantages, but in areas such as chip module selection, voice technology, prompt engineering, intelligent agent development, and communication integration, it is precisely these areas that represent the technical shortcomings of traditional manufacturers. This imbalance in division of labor leads to long POC cycles and high investments for AI toys. If traditional manufacturers want to enter the market, they must build software and hardware teams, facing the pressure of “visible investment thresholds” and “invisible return on investment”.
In terms of production efficiency, traditional manufacturers hold user insights, production lines, and sales channels, but are limited by their technical implementation capabilities, unable to predict what level of intelligent effects can be achieved, nor how much time and money need to be invested. In terms of economic benefits, the toy industry is inherently difficult to predict for blockbuster hits. Even with heavy investments in R&D, success or failure still depends on uncontrollable factors such as market trends and social dissemination.
This also explains why the AI toy market is so attractive—AI toys are often positioned in the mid-to-high-end market, with prices generally above 299 yuan, and even reaching thousands, yet most traditional toy manufacturers find it difficult to get started—many manufacturers we encounter are optimistic about AI transformation, but in the actual promotion process, they find numerous obstacles, often troubled by unclear technical paths and difficulties in controlling implementation effects.
Factories need people who understand content, AI companies need people who understand toys, IP parties need people who understand the industry chain, and channel parties need people who understand product value, AI toys cannot be completed by a single-point company. In the current industrial ecosystem of AI toys, participants often include traditional toy manufacturers, tech toy manufacturers, IP companies, RTC cloud service providers, and intelligent hardware solution providers.
Therefore, the key to breaking the deadlock becomes very clear. Que Hangning emphasized that “Only by lowering the initial investment threshold and risk can more manufacturers dare to enter the field, thus accelerating the intelligent prosperity and upgrading of the entire industry.”
03
AI Capability Foundation: “Emotional Companion Intelligent Agent”
NetEase Cloud Communication launched the “Emotional Companion Intelligent Agent,” building an AI capability foundation by integrating three major modules: communication capabilities, hardware solutions, and emotional computing, achieving a transformation from “providing single-point technology” to “providing overall solutions.”
Traditional toy manufacturers can skip the complex software and hardware selection and integration stages, achieving “plug and play”. By configuring the platform, they can debug dialogue effects in a “what you see is what you get” manner, achieving controllable costs, reducing integration time, and lowering dependence on professional R&D talent, quickly realizing AI intelligent upgrades.
1. The Complete Technical Architecture for “Smooth Dialogue”
NetEase Cloud Communication has built a complete component-based technical architecture. This architecture is based on the WE-CAN global transmission network and the full-terminal RTC SDK, ensuring high-speed and stable signal access. On this basis, the cloud integrates full-link capabilities from ASR multilingual recognition, AI task planning to large model super-human TTS and multimodal processing (covering Prompt, memory, etc.), making every dialogue natural, timely, and full of emotion, achieving low-latency smooth interaction.
By upgrading the cloud RTC engine, NetEase Cloud Communication has significantly reduced dependence on the computing power of terminal devices, allowing mid-to-low-end toys to achieve smooth AI interaction. In terms of hardware solutions, we provide pre-integrated AI engine plug-and-play modules, allowing manufacturers to avoid complex chip selection and underlying integration.
2. The Core Interaction Experience of “Understanding Emotions”
NetEase Cloud Communication is committed to promoting intelligent agents to achieve a fundamental leap from “being able to converse” to “understanding communication.” We have defined the evolution stages of emotional companion intelligent agents: the E0 stage is passive mechanical response scripted interaction, the E1 stage connects to text large models, where the intelligent agent already possesses contextual understanding and short-term memory, progressing to the E2 stage of multimodal emotion recognition, and then to the E3 stage of proactive care and long-term emotional memory, where the intelligent agent gradually establishes attachment, further achieving stable personality and emotional evolution, ultimately reaching the goal of symbiosis.
To this end, we have completed breakthroughs in three core interaction technologies: the “Elegant Interruption” function is based on semantic integrity judgment, allowing AI to choose the most appropriate response timing like a real person; the “Memory System” constructs a three-layer memory framework of short-term, medium-term, and long-term, ensuring that interactions have continuity and growth; the “Story Engine” cleverly solves the initial “ice-breaking” challenge through gamified interactions, gradually establishing a deep emotional connection between users and AI.
This complete technical architecture and interaction system have also been validated through practice. NetEase Cloud Communication helped the client “Baby Story” launch the AI toy “God of Wealth Bi Gan,” completing dialogue effect debugging in just 2 days and engineering design in 2 weeks, quickly realizing the transformation from concept to physical product, allowing traditional toy manufacturers to focus on their strengths in appearance design and user understanding, rather than getting bogged down in technical integration.
04
From Toys to Full-Scene Emotional Companionship
NetEase Cloud Communication insists on the path of “co-development of foundational capabilities and scene applications,” where the foundation is core, and scene capabilities naturally grow on this basis. Even in the specific direction of “emotional companionship,” it can derive applications serving different age groups and needs, such as children’s enlightenment, youth social interaction, and elderly care.
Therefore, the practice in the AI toy field is just the first landing point of the emotional companion AI foundation. Our long-term goal is the continuous iteration of emotional companion intelligent agents from “having” to “excellent,” and through this AI foundation, to co-build with partners across industries, continuously expanding the boundaries of capabilities.
“Supporting the flourishing of ‘Artificial Intelligence +’ across various industries.” This is the goal that NetEase Cloud Communication will continue to uphold in the future.
If you are also an innovator in the AI era, here are some deep thoughts on AI strategy:
1. Carefully assess needs: In capturing AI innovation opportunities, the core challenge lies in accurately discerning the authenticity of needs, finding the true value anchor, avoiding falling into the “illusion on floating sand.”
2. Avoid the coverage area of large models: In capability design, one should avoid the “intermediate layer” that may be quickly disrupted by general large models. Taking the evolution of multimodal large models as an example, it is highly likely that they will improve emotional recognition capabilities in the future; however, the “thousand faces” style of deep personalization architecture will still be a market gap that general technology finds difficult to cover.
3. Recognize the boundaries of AI capabilities: AI excels at handling open-ended and heuristic tasks, but in scenarios requiring precise debugging, major decisions, and complex corrections, “human-machine collaboration” is still necessary. Letting AI handle creativity and generation, while humans handle review and decision-making, is currently the most efficient collaboration model.
4. Choose a differentiated track: Rather than competing with giants in universal scenarios, it is better to focus on niche vertical fields. The key is to accurately calculate the input-output ratio, ensuring that the capabilities accumulated in specific tracks can form barriers and bring continuous benefits.
Scan the QR code below to get a trial of the emotional companion intelligent agent~
About Us
Download Free Resource Materials
