Skip to main content
ukiyo journal - 日本と世界をつなぐ新しいニュースメディア Logo
  • All Articles
  • 🗒️ Register
  • 🔑 Login
    • 日本語
    • 中文
    • Español
    • Français
    • 한국어
    • Deutsch
    • ภาษาไทย
    • हिंदी
Cookie Usage

We use cookies to improve our services and optimize user experience. Privacy Policy and Cookie Policy for more information.

Cookie Settings

You can configure detailed settings for cookie usage.

Essential Cookies

Cookies necessary for basic site functionality. These cannot be disabled.

Analytics Cookies

Cookies used to analyze site usage and improve our services.

Marketing Cookies

Cookies used to display personalized advertisements.

Functional Cookies

Cookies that provide functionality such as user settings and language selection.

Baidu Revolutionizes Search and Video Generation with AI! The Goldmine Baidu Targets Beyond Advertising

Baidu Revolutionizes Search and Video Generation with AI! The Goldmine Baidu Targets Beyond Advertising

2025年07月04日 11:35

1. "Major Overhaul in a Decade"—Search Homepage Goes AI-Native

On the morning of July 3 (Beijing time), Baidu announced the "largest-scale update" of its search engine through its official blog and a briefing at its Beijing headquarters. The search bar has been widened to accommodate "long queries," with voice input and camera input buttons permanently installed. Behind the scenes, the large language model ERNIE 4.5 family interprets the full context and generates answer candidates.


Furthermore, it displays images, videos, shopping, maps, etc., inline in a carousel according to the context—a transformation into a "generative AI-native portal."reuters.comreuters.com


2. MuseSteamer—"From One Image to 10-Second Video" with Three Pricing Tiers

Simultaneously unveiled was the image-to-video model "MuseSteamer." It is structured in three tiers: Turbo, Pro, and Lite, with Turbo generating 1080p videos in as fast as 1.5 seconds, Pro in 4 seconds, and Lite in 10 seconds. All can output up to 10 seconds from a single still image, with test offerings starting for corporate users only. Baidu emphasized that its lip-sync accuracy for Chinese voice is higher than Sora or Gen-2.reuters.comtechinasia.com


3. Why "B2B Only"?—The Realities of Regulations and Monetization

While OpenAI's "Sora" and Runway's "Gen-3 Alpha" open doors to general creators with a monthly subscription model, MuseSteamer is limited to corporate-only beta. The background includes

① Companies that bear copyright clearance obligations are more likely to pass China's cyber regulatory reviews

② High-priced SaaS for corporations can compensate for the decline in advertising revenue

③ Inquiries about output quality directly impact support costs

—among other calculations. Reuters analyzed it as a strategy to reclaim search traffic lost to chatbot platforms like Doubao (ByteDance) and Yuanbao (Tencent).reuters.comreuters.com


4. 48 Hours of Buzz on Social Media—70% Positive, 30% Skeptical

 


  • Positive

    • "Videos with Chinese voice at the touch of a button!" (@hasantoxr)twitter.com

    • "The Chinese company version of Midjourney + Sora, and it's super fast" (@LearnWithBishal)twitter.com

    • AI News Japanese account reported, "Trial registration is open on Esou (Chinese version site) this week."x.com

  • Negative/Concerns

    • On Weibo, posts like "Interest halved if individuals can't access it" and "Videos might be censored with blacked-out words" were observed.

    • In the English-speaking X community, there were remarks like "Only corporate contracts? Without clear API pricing, it won't compete with Sora."

Quantitatively, the "BIDU Bullish" message ratio on Stocktwits jumped by 14 points compared to the previous day, but even within the same thread, cynical voices continued, saying "As long as copyright risks remain, the upside is limited."stocktwits.com


5. Technical Deep Dive

MuseSteamer is a hybrid configuration that merges "Conditioned Latent Diffusion," which extends static image embeddings and diffusion models in the time axis direction, with the ERNIE Vision encoder. The Turbo version generates in 1.5 seconds using 4 GPU A800×8 cards, while the Pro version takes 4 seconds with A100×4 cards—detailed specs were published on a Chinese tech blog. This reportedly reduces lip-sync error (LSE) by 18% compared to OpenAI's Sora.mfro.net


6. Competitive Landscape: Each Company's "V-Gen" Roadmap

CompanyModelRelease ScopeRecommended LengthPrice (Trial)Positioning
BaiduMuseSteamerB2B β10 secondsNot disclosedChinese language specialization
OpenAISoraReserved60 secondsUndecidedGeneral-purpose
ByteDanceMagicVideo 3Consumer-oriented15 secondsMonthly ¥1,980Short video integration
RunwayGen-3 AlphaEarly Access16 secondsMonthly $12Film production-oriented


Baidu aims to differentiate by focusing on "speed and native language voice" rather than length, targeting a "double cropping" strategy of search traffic (advertising) and corporate SaaS (subscription).


7. Market Impact

The day after the announcement, BIDU shares listed in Hong Kong rose by as much as 3.1% from the previous day but ended with only a 0.6% increase. Goldman Sachs maintained its target price citing "uncertainty in the competitive environment." Investors are keenly observing the time lag from "product demo to actual subscription."stocktwits.com


8. Challenges and Prospects

  1. Copyright Clearance—The model that delegates the confirmation of input image rights to companies may face notices equivalent to the U.S. DMCA.

  2. Censorship and Ethics—Services aimed at the domestic Chinese market cannot avoid political filters, raising concerns that the generated results may not meet global market standards.

  3. GPU Costs—The cluster supporting Turbo's inference speed is centered on A800/A100. As ChatGPT's inference optimization progresses, there are concerns about lagging in cost per computation.

Baidu plans to expand the Lite version's API to overseas corporations by the end of the year and is considering a subscription for individual creators by early 2026.


9. Conclusion

The full renewal of MuseSteamer and search AI marks Baidu's "second founding," replacing its existing ad-dependent business with AI SaaS. However, the barriers to monetization and global adaptation are high. Depending on the movements of competitors and changes in the regulatory environment, the challenge could either rewrite search history as a success story or become a short-lived spectacle.


Reference Articles

Baidu Unveils AI-Powered Search Revamp and Image-to-Video Model "MuseSteamer"
Source: https://seekingalpha.com/news/4465258-baidu-shows-off-search-with-ai-revamp-image-to-video-model-musesteamer?utm_source=feed_news_all&utm_medium=referral&feed_item_type=news

← Back to Article List

Contact |  Terms of Service |  Privacy Policy |  Cookie Policy |  Cookie Settings

© Copyright ukiyo journal - 日本と世界をつなぐ新しいニュースメディア All rights reserved.