--- title: "Xinhua Sends an AI-Made Short Video Past 50 Million Views at the Two Sessions" date: 2026-10-29 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-xinhua-ai-news-video language: en --- # Xinhua Sends an AI-Made Short Video Past 50 Million Views at the Two Sessions > Xinhua (新华社) produced "2026, Rivers and Mountains Toward the Future" with AI across the entire workflow, a signal that generative media is moving into mainstream news production. A newsroom editor feeds a large model years of past coverage, asks it to propose a narrative angle, then watches the same system draft the script, generate the visuals, and even write and perform the theme song. The result was not a lab demo. It ran on the home page of China's flagship news agency during the country's biggest political meeting of the year. ## Key takeaways - Xinhua (新华社) New Media Center produced the AI creative short video "2026, Rivers and Mountains Toward the Future" (《2026,山河向未来》) for the 2026 Two Sessions, applying AI across the entire production and distribution workflow. - Xinhua said the piece was picked up by more than 600 media outlets and exceeded 50 million views, topping Douyin and Kuaishou hot-search rankings. - The theme song — lyrics, composition, arrangement and vocals — was created entirely with generative AI tools. - "AI Plus" (人工智能+) appeared in the Government Work Report for a third consecutive year in 2026, and "intelligent agents" (智能体) were written in for the first time. - The experiment exposes the real bottleneck: human judgment, not generation, now decides trust. ## What Xinhua actually did For the 2026 Two Sessions (全国两会), the review of the 15th Five-Year Plan (十五五) outline, Xinhua's (新华社) New Media Center built a short video in which AI touched every stage. In planning, a large model analyzed past Two Sessions coverage to help pick a narrative angle. In scripting, AI generated multiple framework versions from curated source material. In visuals, the team combined several AI tools to render flowing, policy-themed scenes. The theme song — from lyrics and composition to arrangement and singing — was produced entirely by AI. Xinhua (新华社) described it as the first time AI was applied through the entire production and distribution workflow, "full process, full element" (全流程全要素). The agency said the piece was adopted by more than 600 media outlets and surpassed 50 million views, reaching the top of Douyin and Kuaishou hot-search lists. Those distribution figures are Xinhua's own disclosure, relayed by Tencent News and the trade paper China Press, Publication, Radio, Film and Television News (中国新闻出版广电报); we report them as the agency's claims, not as independently audited reach. ## Why this matters beyond one video The 2026 Government Work Report deepened "AI Plus" (人工智能+) for a third straight year and, for the first time, put "intelligent agents" (智能体) on the agenda. Mainstream media read that as permission to move AI from a sidebar experiment into the core production line. Xinhua was not alone: the same reporting season saw robot reporters, virtual hosts, and AI comic explainers across several state and provincial outlets. The through-line is人机协同 — human-machine collaboration — where the machine handles drafting, translation, and multimodal conversion while editors keep the final gate. An industry survey cited by Xinhua found that 96.67 percent of Chinese news organizations said they had already used AI tools by 2025. The question is no longer whether newsrooms use AI, but how they keep the human in the loop when the loop is mostly automated. The same Two Sessions season showed the shape of that loop: robot reporters drafted questions from comment data, virtual hosts read bulletins around the clock, and editors rebuilt their job around verification rather than transcription. None of those roles disappeared; they moved up the value chain toward judgment. What makes the Xinhua experiment a marker rather than a one-off is scale of ambition. A flagship agency putting AI through the entire workflow normalizes the practice for the thousands of smaller outlets that imitate it. When the template spreads, the differentiator is no longer the ability to generate, which everyone will have, but the discipline to verify — and the willingness to tell audiences what was machine-made. ## The trust problem Xinhua (新华社) itself draws the line plainly: no matter how deep AI goes, the final gate must stay with people — fact-checking, value calibration, and publication decisions remain human responsibilities. That is not posture. When a model can draft a polished explainer in minutes, the risk is not bad grammar but confident fabrication at scale, algorithmic bias baked into distribution, and blurred accountability when something is wrong. The agency's own write-up argues for a dual human-machine gatekeeping mechanism and "green algorithms" with explicit responsibility. ## Honest limitations The viewership and pickup figures are Xinhua's own disclosures, not third-party audited numbers; we present them as reported. The piece is a single flagship example, not a measure of routine AI adoption across Chinese newsrooms. We did not independently verify the specific AI tools used in production. Coverage of this experiment comes largely from Chinese media and Xinhua's own published reflections, so non-Chinese readers should treat the workflow description as the agency's account of itself. ## What readers can do now - If you work in news, pilot AI for draft generation and multilingual conversion but keep a named human editor accountable for every published claim. - If you consume AI-produced news, check whether a piece discloses its production method; silence about generation is now a question worth asking. - If you build media AI, design for transparency and provenance first, because trust — not output quality — is the scarce asset once generation is cheap. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-xinhua-ai-news-video Free to quote with attribution and a link to the original.