[Doubao can now watch, listen, and speak simultaneously, and even proactively remind you] According to monitoring by Dongcha Beating, ByteDance's Seed has launched SeedRealtime and rolled it out on the Doubao App. It can process both audio and visuals simultaneously, allowing users to watch, listen, and respond at the same time. Compared to the previous generation Seeduplex, which could only listen and speak, this version introduces real-time visual capabilities. Users can have it focus on the camera to locate objects, check operations, or read text. When a target appears, the scene changes, or the user makes a mistake, it proactively reminds you without needing to wait for a question. In multi-person or noisy environments, it can also determine who is speaking and use facial recognition, voice, and gestures to understand the context. ByteDance claims that issues like interrupting conversations, slow responses, and noise-triggered errors have been reduced by about half compared to traditional stitching solutions. [Original link]