Kling Video (3.0 / Omni / Turbo)

Official Kling AI Open Platform video models (JWT auth, same keys as Kolors VTON).

CLI model	API model_name	Best for
kling-v3	kling-v3	High-quality T2V / I2V, 3–15s, audio
kling-v3-omni	kling-v3-omni	Multimodal refs, elements, Omni workflows
kling-v2-5-turbo	kling-v2-5-turbo	Fast previews

Environment

export KLING_AI_API_KEY=...
export KLING_AI_SECRET_KEY=...
# export KLING_AI_BASE_URL=https://api-singapore.klingai.com

CLI

opentryon video-generate --model kling-v3 \
  --prompt "Slow pan across a couture atelier" \
  --duration 5 --mode pro --aspect-ratio 16:9 --sound on

opentryon video-generate --model kling-v3-omni \
  --prompt "Multi-shot lookbook" --duration 8 --mode pro

opentryon video-generate --model kling-v2-5-turbo \
  --prompt "Quick product spin" --duration 5 --mode std

# Image-to-video
opentryon video-generate --model kling-v3 \
  --image start.jpg --prompt "Subject turns and smiles" --duration 5

Python

from tryon.api.kling_video import KlingVideoAdapter

adapter = KlingVideoAdapter(model="kling-v3")
video = adapter.generate_text_to_video(
    prompt="A model walking a marble corridor",
    duration="5",
    mode="pro",
    sound="on",
)
open("kling.mp4", "wb").write(video)

Endpoints used

• POST /v1/videos/text2video
• POST /v1/videos/image2video
• POST /v1/videos/omni-video (Omni)
• Task polling via GET /v1/videos/{kind}/{task_id}