Baidu shows off search with AI revamp, image-to-video model MuseSteamer

Baidu (NASDAQ:BIDU) confirmed that it has revamped its Search with New AI features in its biggest overhaul in a decade.

The Chinese internet search giant also confirmed the launch of its image-to-video model MuseSteamer.

Baidu said that at the heart of this revamp is the upgraded Smart Search Box that supports long-form queries of over 1,000 Chinese characters, accepts multimedia inputs including image, audio, video, and file, and integrates directly with AI tools for writing, image generation.

Baidu noted that if a user is planning a travel itinerary or trying to recall a vague movie plot, the Smart Search Box can interpret the request and deliver precise, helpful results. For example, users can take a photo of a scenic spot for instant tour commentary, snap a faulty appliance for troubleshooting, or upload a medical report for AI-powered interpretation.

Baidu added that Search’s upgraded Baikan feature restructures search result pages, going beyond traditional links to deliver structured answers that integrate formats including text, images, videos, charts, among others.

With services like AI agents to be integrated, Baikan provides tailored, actionable information — such as organizing product recommendations and even enabling direct purchases from the search page.

The AI Assistant feature now offers video calling, multimodal input (text, voice, image, and files), and outputs including notes, images, and videos, according to the company.

Baidu noted that upgraded Baidu Search also features advanced content creation capabilities, offering a one-stop solution from ideation to video generation to publishing.

Users can create a three-minute AI-generated video from a single sentence, with built-in tools for storyboard editing and custom scene design, the company added.

Baidu also confirmed the launch of MuseSteamer, an image-to-video model enabling video generation featuring synchronized visuals, sound effects, and dialogue. Baidu is integrating MuseSteamer into Baidu Search.

The model can simultaneously generate synchronized visuals, speech, and sound effects in Chinese. It can produce 10-second 1080P cinematic-quality clips from just one image — complete with nuanced facial expressions and camera motion.

The MuseSteamer model family includes Turbo, Lite, Pro, and a full lineup of voiced versions. The Turbo version is currently available for limited-time free public testing on Huixiang, Baidu’s AI video creation platform, with the other versions set to roll out starting in August, the company added.

Baidu was among the first group of Chinese companies whose AI model was approved in China following the launch of OpenAI’s ChatGPT in late 2022 in the U.S. However, the company’s Ernie Bot faces fierce competition in the Chinese market, with tech giants like Alibaba (BABA) and startups such as MiniMax, Baichuan and Zhipu AI also introducing AI models.