Real outputs from the public examples of this model on Replicate, with the prompt and settings that produced them. They were not generated live on this page.
Prompt
First-person POV of a roller coaster plunging into the mouth of an erupting volcano. The track spirals down through rivers of glowing orange lava, sparks and embers flying past the camera. The coaster banks hard around a pillar of molten rock, then launches upward through a vent, bursting out of the volcano's crater into a dazzling sunset sky above the clouds. Wind roaring, riders screaming with excitement, the deep rumble of the volcano.
Bamse, the world's strongest bear, a cartoon brown bear wearing blue overalls and a blue knit cap, lifts an enormous boulder over his head and hurls it across a sunny Swedish meadow. Wildflowers scatter in the wind. His friends, a small rabbit and a turtle with a shell full of gadgets, cheer from behind a wooden fence. Bright, colorful 2D animation style with thick outlines. Cheerful music, a triumphant fanfare, birds chirping.
Cinematic drone footage of a dragon made of living fire soaring over a frozen Nordic fjord at twilight. Its wings trail streams of flame that melt the ice below, creating swirling clouds of steam. The dragon banks sharply and breathes a column of fire across the mountainside, igniting the snow in brilliant orange and gold. Northern lights shimmer green and purple in the sky above. Thunderous wingbeats, crackling fire, hissing steam, a deep otherworldly growl.
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.
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.
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.
Formation & licence
Closed multi-billion-clip corpus combining licensed video, web video and Kuaishou platform data, with dense audio-visual captions; exact size undisclosed.
Licence: Proprietary commercial licence via Kling AI / Kuaishou terms; commercial use on paid plans.
curl https://railwail.com/api/v1/videos/generations \
-H "Authorization: Bearer $RAILWAIL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kling-v3",
"prompt": "A slow drone shot over a misty pine forest at sunrise"
}'
# Response: {"job_id": "...", "status": "queued", ...}
# Poll until status is completed, failed or cancelled:
curl https://railwail.com/api/v1/jobs/JOB_ID \
-H "Authorization: Bearer $RAILWAIL_API_KEY"
import os
import time
import requests
API = "https://railwail.com/api/v1"
headers = {"Authorization": f"Bearer {os.environ['RAILWAIL_API_KEY']}"}
job = requests.post(
f"{API}/videos/generations",
headers=headers,
json={
"model": "kling-v3",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
},
).json()
while True:
status = requests.get(f"{API}/jobs/{job['job_id']}", headers=headers).json()
if status["status"] in ("completed", "failed", "cancelled"):
break
time.sleep(5)
print(status["status"], status.get("output_url"))
const API = "https://railwail.com/api/v1";
const headers = {
Authorization: `Bearer ${process.env.RAILWAIL_API_KEY}`,
"Content-Type": "application/json",
};
const job = await fetch(`${API}/videos/generations`, {
method: "POST",
headers,
body: JSON.stringify({
model: "kling-v3",
prompt: "A slow drone shot over a misty pine forest at sunrise"
}),
}).then((r) => r.json());
let status;
do {
await new Promise((r) => setTimeout(r, 5000));
status = await fetch(`${API}/jobs/${job.job_id}`, { headers }).then((r) => r.json());
} while (!["completed", "failed", "cancelled"].includes(status.status));
console.log(status.status, status.output_url);
