An American director, Charles Curran, after trying it himself, remarked, "This thing might actually disrupt Hollywood." Meanwhile, online users are unleashing their creativity—videos of cats beating up Godzilla, Ip Man fighting Iron Man, and Sun Wukong battling Ultraman are flooding social media feeds.

Recently, ByteDance's latest video generation model, Seedance 2.0, has gone viral not only domestically but also dominated overseas social media platforms, thanks to its multimodal input capabilities and exceptional reference handling. Some have dubbed it "another DeepSeek moment." On February 12, Elon Musk commented on a post about Seedance 2.0 on his platform X, stating, "Model development is moving too fast."

Even before the arrival of the Year of the Horse, Haidian's AI sector is already in full swing, with multiple AI companies intensively releasing new models, creating a scene of rapid expansion. This AI powerhouse is charging into the new year with immense momentum, adding a strong technological vibe to the Spring Festival and stirring up a global AI wave led by China.

**Global Intelligent Creation "Director"**

How powerful is Seedance 2.0? By simply inputting prompts or uploading reference images, it can produce multi-shot videos with complete native audio tracks in under 60 seconds. Supporting four input modalities—image, video, audio, and text—it offers richer expression and greater control over generation. Users can specify the desired visual style with an image, define character actions and camera movements with a video, and convey the intended rhythm and atmosphere with an audio clip. "Prompts" are no longer limited to text, making the creative process more natural, efficient, and akin to being a real "director."

Several American film professionals, after using it, publicly expressed sentiments like "Hollywood is finished." Director Charles Curran created a trailer for his unreleased live-action film using Seedance 2.0, completing 1 minute and 24 seconds of cinematic-quality footage with synchronized audio, multi-camera "shoots," and other effects in just 20 minutes at a cost of $60.

In multi-dimensional evaluations, Seedance 2.0 demonstrates industry-leading performance in areas such as multimodal reference-based generation, complex audio-visual instruction following, complex motion stability, professional cinematography, audio-visual expressiveness, and audiovisual integration.

Simultaneously, ByteDance's Seed team also launched the Seedream 5.0 Lite intelligent image creation model. As a further step toward a unified multimodal model, it begins to "understand" the intent behind user instructions like a human designer, "recognize" patterns in different images, and apply world knowledge to image and text creation.

ByteDance's music streaming app, Soda Music, has also reached 140 million monthly active users, with daily active users firmly holding the third spot in the industry. As the Spring Festival approaches, people switching from intense work modes to slower-paced living and travel have brought another wave of traffic to Soda Music.

**AI Powerhouse in Full Gallop**

A single flower does not make spring, and a single tree does not make a forest. The popular products from Haidian's AI companies are not emerging sporadically but are forming a trend of fierce competition.

On January 27, Moonlight officially released its new generation open-source model, Kimi K2.5. When tackling complex real-world problems, K2.5 is no longer a single expert but functions like an instantly assembled professional team, autonomously coordinating up to 100 specialized avatars with different backgrounds to handle up to 1,500 steps in parallel.

On January 30, Shengshu Technology launched its video large model, Vidu Q3, which possesses integrated "generate-and-it's-a-film" audio-visual capabilities. It combines visuals, sound, and long duration into one, outputting a complete 16-second narrative segment in a single generation, with support for multi-character, multi-language dialogue. Vidu Q3's "director's mindset" intelligently schedules shots, flexibly switching cinematography based on narrative tension, achieving complex transitions from macro to micro within a single generation.

On February 5, Kuaishou's Kolin AI officially launched its 3.0 series models globally, returning the utmost freedom of visual direction to every creator, truly ushering in a "director's era for everyone." The goal is to integrate the traditionally multi-person, multi-revision video production workflow into a more direct and controllable single creative interface, significantly boosting the efficiency of realizing ideas. Starting February 6, Kolin AI officially released its first AI Lunar New Year animated short film series, "Immediate Drama," updated daily on Kuaishou's Star Short Drama and Kolin AI Creative Circle platforms.

On February 11, Zhipu AI officially released its new model, GLM-5, which excels in complex system engineering and long-range agent tasks, achieving state-of-the-art (SOTA) open-source performance. Its user experience in real programming scenarios approaches that of Anthropic's Claude Opus 4.5 model. After its release, GLM-5 ranked in the top three globally and first among open-source models on the authoritative Artificial Analysis leaderboard.

Within Haidian's AI powerhouse, there are many other well-known and frequently used large models familiar to netizens. WPS AI, Xiaoai Assistant AI, Beike Dreamer Large Model, Didi Chuxing Large Model, DeepseekChat... These companies' large models have become valuable assistants in users' work, study, and daily lives.

As of February 14, Haidian District has completed the filing for 128 generative AI services, accounting for 59.5% of the city's total. These services cover government affairs, scientific research, education, culture, tourism, creativity, e-commerce, lifestyle services, and more, showcasing a diverse "general + vertical" layout.

As technology evolves, generative AI is advancing from "content generation" to "solution generation," moving from single-modal to multimodal integration, and from general foundations to vertical specialization, better empowering countless industries.

**"Strongest Brain" Empowers the "AI + Audio-Visual" Industry**

Within just over a month into 2026, the stunning debuts of Seedance 2.0, Kolin AI 3.0 series, and Vidu Q3 have created massive waves in the global "AI + audio-visual" industry. Why Haidian? The answer lies in the fact that Haidian's audio-visual industry and artificial intelligence sector both possess a "strongest brain" and a solid foundation.

As one of the core national hubs for the audio-visual industry, Haidian is home to leading national film and television institutions like China Media Group, China Film Group, Huaxia Film, and the Beijing Film Academy; digital media giants represented by Sina, Sohu, and Jinri Toutiao; video content platforms like Douyin, Kuaishou, Tencent, iQiyi, and Baidu; and quality gaming companies such as Perfect World, Ourpalm, Longchuang Yuedong, and Zhen Entertainment.

Data shows that by 2025, Douyin's daily active users had surpassed 600 million. Kuaishou is a content community with 400 million daily active users; as of December 2025, Kolin AI had over 60 million creators generating more than 600 million videos. Shengshu Technology's Vidu users span over 200 countries and regions worldwide, exceeding 30 million.

The vast amounts of data accumulated by these platform users provide continuous nourishment for the ongoing iteration of large models. The viral hits created by users' inspired ideas, accumulated over time, contribute to the success of these large models. Haidian's "strongest brain" nurtures the robust growth of the "AI + audio-visual" industry. Large models like Seedance 2.0, Kolin AI 3.0, and Vidu Q3, refined through rigorous development, are expanding globally based on their capabilities.

Haidian District is committed to building a national AI audio-visual industry demonstration base. In October 2025, measures to support "AI + Culture," promote high-quality development in the gaming and esports industry, and foster high-quality micro-short drama production in Haidian were released, leveraging the dual-drive advantage of culture and technology to propel Haidian's high-quality economic and social development. Haidian's "AI + audio-visual" industry is ushering in new development opportunities.

**Forging a Global AI Innovation Hub**

With everything prepared, all awaits the first thunderclap of spring. On the eve of the Spring Festival, Haidian's generative AI large models collectively "leveled up," announcing their arrival to the world—a result of sustained accumulation and natural progression.

Haidian boasts the strongest research foundation, the most potent R&D innovation capabilities, and the most active product iteration cycle in China's AI field, having formed a complete industrial ecosystem encompassing "chips - frameworks - large models - applications." Haidian is home to 12,300 AI scholars and over 2,000 AI companies, with a core AI industry scale nearing 360 billion yuan. Leveraging its regional resources, Haidian is striving to create a globally influential AI innovation source and industrial hub, having introduced a series of supportive policies, including measures to lead the development of the general AI industry.

On February 13, the market capitalization of AI chipmaker Cambricon reached 472.6 billion yuan. On January 22, in an internal institutional communication, Loongson Technology indicated it is fully developing its self-developed mid-to-high-end graphics card, the 9A2000. On February 12, Moore Threads first revealed the hardware parameters of its flagship AI training-and-inference integrated full-feature GPU, the MTT S5000. According to industry sources, the MTT S5000's tested performance is comparable to NVIDIA's H100, with some aspects even surpassing the H100 in multimodal large model fine-tuning tasks.

During the "16th Five-Year Plan" period, Haidian District will rely on the Beijing AI Origin Community, the Zhongguancun AI North Latitude Community, and the Beijing-Zhangjiakou Railway Heritage Park AI Innovation Belt to build a benchmark cluster for Beijing's AI industry.

The origin of this current wave of Chinese AI breakthroughs lies in Haidian, supported by the entire industrial chain and Haidian's science and technology innovation ecosystem. The leading Haidian AI powerhouse vividly exemplifies Haidian's determination to be the "leading goose" in the AI race. In the Year of the Horse, Haidian's AI industry is poised to advance smoothly and achieve great success.