kling-3-0

Kling 3.0 video generation accepts text prompts or HTTPS reference images to create cinematic videos. You select a tier and mode, submit JSON inputs including prompt, duration, and aspect ratio, then the CLI polls and downloads the final video.

Yes, Kling 3.0 image-to-video animation uses an HTTPS image_url as the starting frame and an optional tail_image_url to lock the ending frame. This ensures your generated video starts and ends with your specified visual inputs.

Kling 3.0 supports synchronized in-pass audio generation including dialogue, ambience, and music. You enable this by setting generate_audio in your JSON inputs, with tier-aware pricing behavior applied to the Standard, Pro, or 4K tiers.

Kling 3.0 maintains consistent character identity across multiple shots using multi-segment cinematic prompting. This scene-by-scene prompt structure allows you to define consistent characters and narratives throughout the generated video.

RunComfy Kling 3.0 endpoint routing requires JSON inputs containing your prompt, image_url for image-to-video, duration, aspect_ratio, and generate_audio. Optional fields include seed, negative_prompt, and tail_image_url for ending frame control.