miHoYo Launches AI Companion App BSide: Olivia Lin on Steam Early Access
Chinese gaming firm miHoYo has launched its AI companion app, BSide: Olivia Lin, in Early Access on Steam. The release signifies miHoYo's expansion beyond gaming into AI-powered digital companionship.

Chinese gaming firm miHoYo released its AI companion app, BSide: Olivia Lin, to Steam Early Access on Monday. Listed as a free-to-play utility rather than a game, the release marks miHoYo's latest move to expand beyond interactive entertainment into AI-powered digital companionship.
The app centers on Olivia Lin, a virtual character portrayed as a Shanghai student majoring in piano with psychology as a minor. Her profile states an enjoyment of vinyl records, classic films, and rainy days, with a focus on exploring the relationship between music and memory. The character initially gained traction on Bilibili in August 2025 through motion-captured piano performance videos.
Unlike miHoYo's typical game releases, BSide: Olivia Lin functions as a lightweight companion app. It combines AI interaction, music creation tools, and desktop customization. Users can watch Olivia perform piano pieces, or musicians can generate animated performance videos from their own MIDI files using AI. The app also includes a letter-based AI chat system where Olivia responds in character to user inputs.
Early Access users have noted a discrepancy between expectations and the product, with many downloading it assuming it was a new game. Feedback has also cited formulaic AI responses, bugs with audio-video synchronization in generated videos, and a cumbersome account linking process required for login. Some users have questioned the app's value as a standalone product versus a data-gathering tool for AI training.
miHoYo states the project is under active development, with Early Access intended to gather feedback for future feature expansion and refinement. The app has already surpassed 100,000 free downloads on Steam. Its success in the competitive AI companion market will likely depend on its ability to establish a differentiated value proposition in future iterations.