A studio in Hangzhou used to charge clients for a three-day shoot to produce a 10-second product clip. This quarter the same client asked for 200 language variants and routed the budget to a text-to-video (文生视频) account instead. The move from camera crews to prompt boxes is no longer a forecast — Kuaishou's (快手) latest earnings just attached a number to it.
Key takeaways
- Kling AI (可灵) booked RMB 340 million (≈ US$47M / HK$370M) in Q4 2025 revenue, according to Kuaishou's full-year results released in March 2026.
- December 2025 alone exceeded US$20 million in monthly revenue, lifting the annualized run-rate (ARR) to US$240 million.
- By end-2025 the service reported 60 million-plus global users, 600 million-plus generated videos, and 30,000-plus enterprise and developer API customers.
- The Kling family shipped O1, 2.6, and 3.0 inside a single quarter, with 2.6 introducing simultaneous audio-plus-video output ("音画同出").
A model that became a line item
Kuaishou built Kling as an in-house answer to the global text-to-video race, and until 2025 it was mostly discussed as a showcase feature inside a short-video app. The Q4 numbers change that framing. RMB 340 million in three months is small next to Kuaishou's RMB 142.8 billion (≈ US$19.8B) full-year revenue, but the growth curve is the story: a single-month figure above US$20 million and a US$240 million run-rate imply the product is compounding, not merely piloting.
The company's broader figures give context. In 2025, Kuaishou's adjusted net profit reached RMB 20.6 billion (≈ US$2.9B), and its app averaged 408 million daily active users in Q4. Kuaishou also said AI-produced (生成式内容) ad creatives consumed roughly RMB 4 billion (≈ US$550M) across its marketing system last year — a sign that generative media is already embedded in how the platform sells ads, not just how it makes demos.
What changed in the product
Three model iterations landed in one quarter. Kling O1 was positioned as a unified multimodal workspace, folding text, image, video, and subject references into one engine. Kling 2.6 added "音画同出" — generating picture and synchronized sound together rather than dubbing silent footage afterward. Kling 3.0, released in February 2026, pushed an "All-in-One" concept toward professional scheduling of shots, per the company's description.
The audio-plus-video step matters more than it sounds. Most text-to-video tools still produce silent clips that need a separate voiceover pass. Bundling sound changes the workflow for advertisers and educators who need a finished, speakable asset, not a pretty loop.
Why China's video models are monetizing first
Kling's run-rate is not just a Kuaishou story; it reflects a wider pattern. Chinese text-to-video products are being wired directly into commerce, short-drama (短剧), and ad pipelines, where volume and iteration speed translate into paying use almost immediately. A 10-second clip that used to need a crew now needs a prompt and a subscription.
Kuaishou's own count of 30,000-plus enterprise and developer API customers suggests the demand is not only consumer curiosity. When studios, game teams, and e-commerce sellers build Kling calls into their production stacks, the model becomes infrastructure rather than a toy.
The competitive read
Western text-to-video leaders have chased cinematic quality and viral demos; Chinese leaders are chasing throughput and integration. Kling's edge, by Kuaishou's account, is less about winning a single beauty contest and more about being the default engine inside a massive content economy. Whether that lead holds depends on model quality, which remains hard to compare across labs, but the commercial footprint is already visible in the income statement.
What the run-rate doesn't say
A US$240 million run-rate is a snapshot, not a trajectory. It tells you December was strong; it does not tell you whether January held. Kuaishou has not broken out retention, churn, or average revenue per enterprise customer for Kling, so we cannot tell how much of the number is a few large accounts versus a broad base. That distinction matters: a base of many small studios is more durable than dependence on a handful of whales. Investors and buyers should watch the next two quarterly disclosures for whether the run-rate compounds or flatlines.
Honest limitations
All revenue and user figures in this article come from Kuaishou's 2025 full-year earnings as reported by Chinese financial media in March 2026; we did not independently audit them. "ARR" is a run-rate based on December 2025 monthly revenue and is not a guaranteed forward number. The RMB-to-USD/HKD conversions use approximate rates (1 USD ≈ 7.2 RMB; 1 RMB ≈ 1.09 HKD) and will shift with exchange rates. We did not find third-party benchmarks in this pass that independently confirm Kling's quality lead over rivals such as Seedance (即梦) or international models. Kuaishou's claims about model capabilities are product descriptions, not independently verified benchmark results.
What readers can do now
- If you produce ad or e-commerce video, test Kling (可灵) against a real shoot on a single 10-second brief and compare cost and turnaround, not just visual polish.
- Developers should review Kling's API tiers if you need bulk, variant-heavy video generation for storefronts or training content.
- Track the ARR trajectory across the next two Kuaishou quarters before treating text-to-video as a settled line item rather than a fast-growing one.