From GIF Maker to AI Pioneer: How Kuaishou Birthed Kling, a Leading Video Generation Model

Kuaishou's Evolution: From GIF Tool to Short Video Giant

Founded in March 2011, Kuaishou originally emerged as a mobile GIF creation tool. In 2012, under founder Cheng Yixiao, Kuaishou pivoted toward a short-video sharing platform, securing substantial Series A funding and rapidly scaling to millions of daily active users. By 2014, it formally adopted the "Kuaishou" brand and cemented itself as a populist video-sharing community.

Over subsequent years, strategic investments from tech giants like Tencent significantly boosted Kuaishou’s user base and technological capabilities, culminating in a successful IPO on the Hong Kong Stock Exchange in February 2021. As of 2024, Kuaishou has amassed over 900 million daily active users in China, becoming a top-tier platform integrating short videos, live streaming, and e-commerce.

Why Kuaishou Launched Kling

The birth of Kling, Kuaishou's proprietary AI video generation model, was driven by practical business needs and strategic foresight. Given its massive community of predominantly amateur content creators, Kuaishou faced persistent challenges in content quality and creation scalability. Generative AI video technologies emerged as a compelling solution, significantly lowering creative barriers and enriching content diversity.

Cheng Yixiao, Kuaishou's CEO, envisioned Kling as a tool enabling anyone to easily create high-quality video content from simple textual or image prompts, empowering vast user creativity and enhancing platform engagement.

Kling’s Rapid Rise and Milestones

Launching Kling 1.0 (June 2024)

Kling launched publicly in June 2024, swiftly integrating into Kuaishou’s ecosystem through its video-editing app, Kuaiying. The first-generation Kling model supported creating two-minute-long, 1080p videos directly from text prompts—significantly ahead of many international competitors, including OpenAI's Sora.

Continuous Innovation (Late 2024)

Post-launch, Kling underwent rapid iterations, introducing advanced features like image-to-video transformations, video continuation, virtual try-ons, and lip-sync capabilities. The accessible distribution via multiple platforms (apps, web, mini-programs) spurred exponential user growth, reaching millions within months.

Breakthrough with Kling 1.6 Pro (December 2024)

Kling 1.6 introduced a high-quality Pro mode, significantly enhancing video fidelity and realism. This model achieved the top spot globally in Arena ELO benchmarking tests by Artificial Analysis, outperforming Google's Veo 2 and Pika Art.

Entering Kling 2.0 (April 2025)

By April 2025, Kling unveiled its advanced 2.0 version, alongside Ketu 2.0 (image generation). Kling 2.0 delivered superior semantic coherence, visual aesthetics, and dynamic realism, continuing its market-leading position. Internal testing showed Kling 2.0 outperforming competitors like Google's Veo 2 and Runway's Gen-4, particularly in video clarity and motion fluidity.

Comparing Kling with Global Peers

Kling’s architecture, similar to OpenAI's DiT (Diffusion in Time) model, utilizes advanced transformer-based diffusion models optimized for high realism and detailed motion. In contrast, Google’s Veo 2 emphasizes ultra-high-resolution video generation (up to 4K) with advanced physics simulations. OpenAI’s Sora, a pioneer in the field, remains limited by shorter video lengths and strict regional availability. Pika Labs distinguishes itself through open community testing and creative editing tools but trails behind Kling and Veo in overall video quality.

ModelTechnical ApproachUser AccessibilityPerformanceKling (Kuaishou)Diffusion + Transformer (DiT), Flow-based acceleration, High realismMass-market, fully integrated in Kuaishou apps, millions of usersArena ELO #1 globally, realistic 1080p videos up to 2 minsOpenAI SoraDiffusion + Transformer, limited creative and complex scenariosLimited public availability, subscription-basedShorter videos (20 secs public), high-quality visuals but limited regional accessGoogle Veo 2Advanced diffusion and transformer modeling, ultra-high definition, motion physicsEnterprise-focused, cloud API availabilityUp to 4K videos, high realism, second to Kling in global benchmarksPika LabsLatent Diffusion, varied creative styles, extensive editing toolsOpen global testing, community-drivenShorter, less detailed videos but extensive creative and editing options