News Lead
Recently in Beijing time, Kuaishou Technology's Kling AI officially released version 1.5, marking a major upgrade that elevates AI video generation capabilities to new heights. The new Kling supports 1080p HD resolution video generation with a maximum duration of 2 minutes, featuring dynamic consistency and motion realism comparable to OpenAI's Sora model. Within just days of release, engagement on X platform (formerly Twitter) has exceeded 300,000, with Chinese users praising it as "China's Sora has arrived," signaling the accelerated transformation of AI video tools from professional barriers to mass applications.
Background
Kling AI is Kuaishou's first text-to-video large model launched in July 2023, which has stood out since its inception with high-quality output and Chinese optimization. The initial 1.0 version could already generate 5-second HD videos, quickly accumulating millions of users and ranking among the top in international benchmarks. Entering 2024, competition in the AI video field has intensified, with OpenAI's Sora stunning the industry with its realistic physics simulation, while models like Runway and Pika have also iterated rapidly. The Kuaishou Kling team seized the opportunity and, after months of optimization, achieved multiple technological leaps in version 1.5. This is not only a continuation of Kuaishou's AI strategy but also a confident work by local enterprises benchmarking against the world's top models.
As a short video giant, Kuaishou possesses massive user data and computing resources, with its AI layout initiated years ago. Kling's iteration benefits from self-developed diffusion models and Transformer architecture optimization, with training data covering hundreds of millions of video clips, ensuring the model's precise response to Chinese prompts. This upgrade comes at a time when the global AI video market is exploding, with user demand shifting from static images to dynamic content.
Core Content
The core highlights of Kling 1.5 lie in multiple technological breakthroughs. First, the resolution has been upgraded to 1080p (1920x1080), with sharp image details, realistic color reproduction, and support for various aspect ratio customizations. Compared to the previous version, clarity has improved by over 30%, suitable for professional editing and social sharing. Second, the duration has been extended from 10 seconds to 2 minutes (128 frames), meaning users can generate complete short films rather than fragmented clips. Official demonstration videos show that a simple prompt like "horses running under the sunset, camera pulling back" can output smooth long sequences without obvious stuttering or distortion.
Dynamic consistency is another killer feature. Kling 1.5 introduces an advanced physics simulation engine, with object motion following real gravity and inertia laws. For example, when generating a "basketball player dunking" scene, the ball trajectory, fabric folds, and sweat splashes are all highly realistic. Officials claim that its 3D spatiotemporal attention mechanism ensures stable character postures in long sequences, avoiding the "drift" issues of early Sora versions. Additionally, prompt comprehension capabilities have been enhanced, supporting complex multimodal inputs such as "combining photos to generate animations," with Chinese natural language processing accuracy reaching over 95%.
In terms of technical details, Kling 1.5 adopts a DiT (Diffusion Transformer) architecture with parameter scale exceeding tens of billions, and inference speed optimized to 30 frames per second. Kuaishou provides free trial quotas, with 10 generations per day, and unlimited output for paid versions, with extremely low barriers. Compared to Sora's invitation-only system, Kling focuses more on accessibility.
Various Perspectives
User feedback has been enthusiastic. On X platform, @AI_Observer posted: "Kling 1.5's 2-minute HD videos are amazing, dynamic consistency completely beats Runway, Chinese prompts work perfectly!" The interaction exceeded 50,000, with most reposts from content creators. Another content creator @VideoAI_fan stated: "From 5 seconds to 2 minutes, editing workload halved, domestic AI finally catching up with Sora's pace." Negative feedback is minimal, mainly focusing on queue times during peak hours.
"Kling 1.5's motion coherence and physical realism have approached Sora's level, especially with Chinese optimization making local users feel at home." — Lee Kai-fu (Founder of Innovation Works), X platform comment.
Industry experts also highly recognize it. Tsinghua University Professor Zhang Yaqin pointed out in an interview: "Kling's spatiotemporal modeling innovation is commendable, proving China's catching-up speed in AI video generation." Former OpenAI researcher @SoraDev anonymously responded: "Not bad, but Sora has richer training data, still gaps in long videos." Kuaishou CTO Yu Haifeng emphasized at the launch: "We focus on practicality, not showing off."
Impact Analysis
The release of Kling 1.5 will profoundly reshape the AI video ecosystem. First, it promotes mass adoption. Previously, video generation required professional equipment, now it only needs a mobile browser for creation, driving innovation in short videos, advertising, and e-commerce live streaming. User base is expected to exceed 10 million within the year, boosting Kuaishou's content ecosystem.
Second, it stimulates industry competition. While Sora leads, its closed access limits adoption; Kling's open-source tendency (partial weight disclosure) and low-price strategy will lower barriers, stimulating iterations from Runway, Luma, and others. The global market size is expected to reach tens of billions of dollars by 2025, with China's share significantly increasing.
Challenges remain: high computing power demands test infrastructure, copyright and ethical issues need vigilance, such as deepfake video generation risks. At the regulatory level, China's Ministry of Industry and Information Technology is formulating AI content standards, and Kuaishou has built-in watermarking mechanisms. Meanwhile, Kling strengthens local advantages, potentially accelerating "Belt and Road" AI export.
In the long term, this upgrade marks text-to-video entering the "HD long-duration" era, with multimodal integration (such as voice, 3D) on the horizon, and the creator economy set to explode.
Conclusion
Kuaishou Kling 1.5 is not only a technological milestone but also a vivid practice of AI democratization. With HD 2-minute output and Sora-level consistency, it has won global attention. In the future, as models iterate, AI video will transform from tools to companions, restructuring human expression. Keep an eye on Kling and await the next surprise.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接