--- title: "Wondershare's Tianmu 2.0 lands above OpenAI's Sora on a Chinese text-to-video leaderboard" date: 2026-11-11 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-wondershare-tianmu-2 language: en --- # Wondershare's Tianmu 2.0 lands above OpenAI's Sora on a Chinese text-to-video leaderboard > At HDC 2025, Wondershare launched Tianmu 2.0, an audio-video multimodal large model built with Huawei Cloud's Pangu, plus a creator platform and an API engine. A filmmaker in Chengdu exports a 30-second ad in the morning, then watches a colleague in Berlin regenerate the same scene with different camera moves by lunch. The gap between a professional studio and a lone creator used to be measured in years of training and racks of hardware. A Chinese software maker is betting that a single model can erase most of it. ## Key takeaways - Wondershare (万兴科技, 300624.SZ) unveiled Tianmu 2.0 (天幕 2.0), an audio-video multimodal large model (大模型), at Huawei's HDC 2025 on June 22, 2025, built on Huawei Cloud's Pangu multimodal model. - On SuperCLUE's July 2025 text-to-video ranking, the Tianmu 2.0 base (Pangu-T2V) placed fourth in China and scored above OpenAI's Sora. - Alongside the model, Wondershare launched a creator platform (tomoviee.cn), a mobile app, and a creation engine with open APIs and MCP support. - The company's 2024 revenue was RMB 1.44 billion (≈ US$200M / HK$1.57B), with overseas sales near 90%, and its AI-native app revenue grew past RMB 67M (≈ US$9.3M / HK$73M), up over 100%. ### A model born from a partnership Wondershare, a Shenzhen-listed digital-creative software group best known overseas for Filmora and PDFelement, has spent two years pushing its products toward generation. Tianmu 1.0 arrived in early 2024 as what the company called China's first audio-video multimodal creation model, and it had already been put to work in short dramas (短剧), film, e-commerce, and finance. Version 2.0 is different in scale: it fuses Huawei Cloud's Pangu multimodal backbone with Wondershare's own library of creative tools and what the company describes as 1.5 billion cumulative end users across its product portfolio. At the HDC 2025 Pangu summit, chairman Wu Taibing (吴太兵) framed the release around "creative equality" — the idea that anyone, not just studios, should be able to produce polished video. The pitch is less about a single magic button and more about a full pipeline: text-to-video, image-to-video, and video continuation, plus image generation, music, sound effects, and voice synthesis. ### What actually shipped Three things went live or were promised at the launch: - **Tianmu Creation Square (万兴天幕创作广场, tomoviee.cn)** — a browser-based studio aimed at professional creators, now in public beta. - **Tianmu AI App (万兴天幕AI App)** — a low-barrier mobile tool for everyday creators, announced as coming soon. - **Tianmu Creation Engine (万兴天幕创作引擎)** — an API and toolset for developers and enterprises, with support for the MCP protocol, so outside software can plug the model in. Wondershare and Huawei Cloud also opened a joint "AI video large-model lab," and the release took home Huawei's Pangu Pioneer Award. The package reads like a deliberate move from selling creative software to renting creative infrastructure. ### The benchmark that matters Independent validation came from SuperCLUE, a Chinese large-model benchmark organization. In its July 2025 text-to-video global ranking, the Tianmu 2.0 base model — built on Huawei's Pangu-T2V — ranked fourth among domestic models and posted a composite score above OpenAI's Sora. SuperCLUE placed it in the "excellent leader" quadrant, citing video quality, scene construction, and dynamic motion, with noted strength in Chinese cultural contexts. Benchmark scores are not the same as real-world quality, and leaderboards shift month to month. But the result is notable because it ties a Chinese consumer-software brand to a foundation model from a major cloud provider, rather than to a model the brand trained alone. ### The business behind the model Wondershare's financials show why a creative-software firm is racing into generation. Its 2024 annual report, filed with the Shenzhen exchange, reported revenue of RMB 1.44 billion (≈ US$200M / HK$1.57B), with overseas markets contributing about 90% of sales. Net profit was negative for the year, weighed down by R&D, sales spending, and a goodwill impairment — a reminder that the generative push is expensive. The generative bet is showing early traction: AI-native application revenue reached roughly RMB 67M (≈ US$9.3M / HK$73M) in 2024, more than double the prior year, and the company said its servers handled over 500 million AI calls. Those are still small against total revenue, but they are the fastest-growing line, and the company reported its AI material library was exported more than 14 million times, up over 700%. ### Why it matters for China's AIGC field Most Chinese text-to-video attention goes to the headline consumer apps. Wondershare's path is the unglamorous one: a toolmaker wrapping a cloud provider's foundation model into editing, effects, and an API, then selling it to creators who already pay for software. If that works, it shows a route for China's many vertical software firms to bolt generation onto existing products without training a frontier model from scratch. The Huawei partnership is the strategic core. By leaning on Pangu for the heavy lifting, Wondershare avoids a multi-year, multi-billion-renminbi race to train its own base model, and instead competes on the creative layer where it already has distribution and user trust. That "model plus vertical app" pattern is becoming the default for Chinese software firms that cannot outspend the hyperscalers. ## Honest limitations This article relies on Wondershare's own announcements, Shenzhen Stock Exchange filings, and Chinese tech-media reports; it does not include independent hands-on testing of Tianmu 2.0's output quality. The "1.5 billion users" figure is the company's cumulative claim across all products, not active paid users. SuperCLUE rankings reflect one organization's methodology at a point in time and should not be read as a guarantee of user experience. We did not verify pricing, exact API throughput, or the mobile app's launch date, which had not been fixed at publication. Overseas revenue share is drawn from the company's 2024 reporting and may shift quarterly. ## What readers can do now - If you edit video for social or e-commerce, trial the Tianmu Creation Square public beta and compare its text-to-video and image-to-video against your current toolchain on one real project. - Developers should check the Tianmu Creation Engine API and MCP support if you need video or audio generation inside an existing product without standing up your own model. - Investors tracking China's AIGC shift should watch whether Wondershare's AI-native revenue keeps compounding past the RMB 67M base and whether overseas share holds above 85%. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-wondershare-tianmu-2 Free to quote with attribution and a link to the original.