Baidu has made headlines with the unveiling of its latest artificial intelligence (AI) video generation model, named MuseStreamer, on Wednesday. This new model is notable for its ability to incorporate Chinese audio into the videos it produces, positioning it as the second AI model of its kind after Google’s Veo 3. Baidu asserts that MuseStreamer is the first AI model globally to support native Chinese audio generation. Alongside this launch, the company introduced a new video creation platform called HuiXiang. However, it is important to note that both MuseStreamer and HuiXiang are currently unavailable outside of China.
Baidu’s MuseStreamer Can Reportedly Generate Chinese Audio
The landscape of AI video generation has undergone significant transformations over the past two years. Initial models faced challenges in generating realistic human figures or accurately representing physics and motion. Yet, a key area that many AI developers avoided was creating videos that offered native audio support.
At the Google I/O event in 2025, Google became the first technology company to launch a model with such capabilities, introducing Veo 3. This innovation garnered considerable attention, overshadowing its competitor, OpenAI’s Sora. Google has recently made Veo 3 accessible across the 154 countries where the Gemini app operates, underscoring its commitment to this tool.
In a recent report by Tech in Asia (via AI Base), Baidu has joined the competition with its MuseStreamer model. This model reportedly generates videos with Chinese audio, a feature that sets it apart, as Veo 3 is limited to English audio generation.
According to Baidu, MuseStreamer is capable of producing dialogues that sync perfectly with its videos, and it can also include sound effects and ambient noises. The company claims that the model achieved an impressive score of 89.38 percent on the VBench I2V benchmark, placing it at the forefront of this technology. Baidu is marketing the LLM as a tool for consumers aiming to create engaging content.
In addition to the AI model, Baidu has launched the HuiXiang platform, which serves as an interface for users to input prompts and generate videos. The platform currently allows for the creation of 10-second videos at a 1080p resolution. In contrast, Veo 3 is limited to generating eight-second videos. Specific details regarding the default aspect ratio of the generated videos, as well as the potential for producing videos in various aspect ratios, remain unclear.
