Kling v3

New
Popular
Replicate
Video Generation

Cinematic video up to 15s with multi-shot and native audio

Queue video with Kling v3
Video generation runs asynchronously — we'll queue a job and you can track it in your history.
Sign in to try this model with €5 free credits.
Sign in
Generates as an async job — typically 30 s to 2 min.
TL;DR·Last updated March 25, 2026

Kling v3 is video generation AI model from Replicate, priced at €0.000 per 1M input tokens with a unknown context window.

Try Kling v3

Image References

Sign in to generate — 50 free credits on sign-up

Examples

See what Kling v3 can generate

0:10

Action

"Parkour athlete on rooftops at sunset"

Pricing

Price per Generation
Per generation€2.00

API Integration

Use our OpenAI-compatible API to integrate Kling v3 into your application.

Install
npm install railwail
JavaScript / TypeScript
import railwail from "railwail";

const rw = railwail("YOUR_API_KEY");

// Simple — just pass a string
const reply = await rw.run("kling-v3", "Hello! What can you do?");
console.log(reply);

// With message history
const reply2 = await rw.run("kling-v3", [
  { role: "system", content: "You are a helpful assistant." },
  { role: "user", content: "Explain quantum computing simply." },
]);
console.log(reply2);

// Full response with usage info
const res = await rw.chat("kling-v3", [
  { role: "user", content: "Hello!" },
], { temperature: 0.7, max_tokens: 500 });
console.log(res.choices[0].message.content);
console.log(res.usage);
Specifications
Price
€2.00
Avg. latency
120.0s
Est. duration
2min
Developer
Replicate
Category
Video Generation
Supported Formats
mp4
Tags
popular
audio
i2v

Deep dive — Kuaishou Technology's Kling v3

About Kuaishou Technology
Founded 2011 · Beijing, China

Kuaishou Technology, founded in 2011 by Su Hua and Cheng Yixiao, is one of China's two dominant short-video platforms, with ~700M monthly active users in 2025. Its KLING AI team shipped Kling 1.0 in June 2024, Kling 1.5 and 1.6 (late 2024), Kling 2.0 (Q1 2025) and Kling v3 (mid-2025). Kling v3 introduced native synchronized audio (sound effects and dialogue), 1080p+ at 30 fps, much improved motion physics, and the Omni multimodal-conditioning variant. The model is exposed via klingai.com, the Kling mobile app and a commercial API and is widely used by Chinese ad agencies, music labels and social-media creators.

Visit Kuaishou Technology →
Architecture
Diffusion Transformer (DiT) with 3D spatio-temporal VAE and joint audio-video diffusion

Kling v3 is a closed-source video diffusion model built on Kuaishou's third-generation Diffusion Transformer architecture. The denoiser operates on a 3D spatio-temporal latent produced by a high-compression VAE and uses full 3D spatio-temporal attention rather than factorised attention. v3's headline change is a joint audio-video diffusion module that generates synchronized soundtrack (ambient sound, footsteps, music swells and short dialogue) aligned to the visual track. Text conditioning uses a bilingual LLM encoder with extended context for very long prompts; image conditioning supports first-frame, end-frame and subject-reference modes. Training uses a curated multi-billion-clip corpus (including platform data) with synthetic dense audio-visual captions; post-training combines reward-model alignment for visual aesthetics, prompt fidelity and audio-video sync.

Parameters
Undisclosed
Context
unknown
What it can do
  • Text-to-video, image-to-video and subject-reference modes
  • 1080p / 30 fps clips up to 10 seconds natively, extendable beyond 1 minute
  • Synchronized native audio (ambient sound, simple dialogue, music swells)
  • Director-style camera control (lenses, rigs, moves)
  • Strong physics on water, fabric, hair and crowd dynamics
  • Bilingual Chinese/English prompts with extended context
  • Lip-sync and motion-brush in Kling editor
  • Commercial API and Kling editor integration
  • Best for: cinematic shorts, ads with audio, music videos, character-driven content.
Training & License

Closed multi-billion-clip corpus combining licensed video, web video and Kuaishou platform data, with dense audio-visual captions; exact size undisclosed.

License: Proprietary commercial licence via Kling AI / Kuaishou terms; commercial use on paid plans.

Known limitations
  • 10-second native clip limit
  • Audio quality is short-form (limited length and music sophistication)
  • Heavy political and public-figure moderation
  • Closed model with no public technical paper
  • Output queues during peak hours

Frequently asked questions

Start using Kling v3 today

Get started with free credits. No credit card required. Access Kling v3 and 100+ other models through a single API.