Kling Video Generation API

Kling covers the full video pipeline: text-to-video, motion control, avatar creation, and lip sync. All models accessible through the Runcrate video API.

Available models

Model	Type	Strengths
Kling v3	Text/Image-to-video	Best human motion, cinematic, 3-15s
Kling v2.6	Text/Image-to-video	Fast, cost-effective, 3-10s
Kling v3 Motion Control	Controlled video	Camera path and subject motion
Kling Avatar v2	Avatar	Talking head from a single photo
Kling Lip Sync	Audio-driven	Sync lips to any audio track

Text-to-video

from runcrate import Runcrate

client = Runcrate(api_key="rc_live_YOUR_API_KEY")
job = client.models.generate_video_and_save(
    "street-scene.mp4",
    model="kuaishou/kling-v3-video",
    prompt="Person walking through rainy Tokyo at night, neon reflections on wet pavement",
    duration=8,
    on_status=lambda j: print(f"  {j.status}..."),
)

import Runcrate from '@runcrate/sdk';

const rc = new Runcrate({ apiKey: 'rc_live_YOUR_API_KEY' });
const job = await rc.models.generateVideo({
  model: 'kuaishou/kling-v3-video',
  prompt: 'Person walking through rainy Tokyo at night, neon reflections',
  duration: 8,
});
let status = job;
while (status.status !== 'completed' && status.status !== 'failed') {
  await new Promise(r => setTimeout(r, 5000));
  status = await rc.models.getVideoStatus(job.id);
}
await Bun.write('street-scene.mp4', await rc.models.downloadVideo(job.id));

All examples below reuse the same client.

Image-to-video

client.models.generate_video_and_save(
    "product-spin.mp4", model="kuaishou/kling-v3-video",
    prompt="Slowly rotate the product 360 degrees on a turntable",
    image="./product-photo.png", duration=6,
    on_status=lambda j: print(f"  {j.status}..."),
)

Motion control

client.models.generate_video_and_save(
    "dolly-zoom.mp4", model="kuaishou/kling-v3-motion-control",
    prompt="A castle on a hilltop at sunset, dramatic clouds",
    camera_motion="dolly_zoom_in", duration=6,
    on_status=lambda j: print(f"  {j.status}..."),
)

Avatar generation

Create a talking-head video from a single reference photo:

client.models.generate_video_and_save(
    "avatar-intro.mp4", model="kuaishou/kling-avatar-v2",
    prompt="Person introduces themselves and waves, friendly, office background",
    image="./headshot.jpg", duration=8,
    on_status=lambda j: print(f"  {j.status}..."),
)

Lip sync

client.models.generate_video_and_save(
    "synced-speaker.mp4", model="kuaishou/kling-lip-sync",
    prompt="Natural lip sync with subtle head movements",
    image="./speaker.jpg", audio="./narration.mp3",
    on_status=lambda j: print(f"  {j.status}..."),
)

Tips

v3 vs v2.6. v3 has better quality and human motion. v2.6 is faster and cheaper.
Duration affects cost. Start with 3-4s clips to test prompts.
Motion control. dolly_zoom_in, dolly_zoom_out, pan_left, pan_right, tilt_up, tilt_down, orbit.

Inference Examples

Compute Examples

Storage Examples

Kling Video Generation API

Available models

Text-to-video

Image-to-video

Motion control

Avatar generation

Lip sync

Tips

​Available models

​Text-to-video

​Image-to-video

​Motion control

​Avatar generation

​Lip sync

​Tips

Available models

Text-to-video

Image-to-video

Motion control

Avatar generation

Lip sync

Tips