Kling O1: Cinematic Image-to-Video Playground & API | RunComfy

The Kling O1 model on RunComfy is specialized for high-fidelity Image-to-Video generation. Unlike standard video models, this Kling O1 workflow allows you to define both a starting frame and an ending frame (using start_image_url and end_image_url), enabling precise control over the video's narrative and motion trajectory.

Commercial usage of Kling O1 depends on Kuaishou's specific licensing terms, as they are the original model creators. While RunComfy provides the infrastructure and API to run the model, you must ensure your project complies with the official Kling O1 usage policy regarding commercial rights and attribution.

To ensure successful generation with Kling O1, your input images must be under 10 MB in file size. Additionally, both width and height must exceed 300 pixels, and the aspect ratio must fall between 0.40 and 2.50. Images outside these Kling O1 specifications may result in validation errors.

Video generation with Kling O1 is computationally intensive. While generation time varies based on the selected duration (5s or 10s), RunComfy’s infrastructure manages concurrency automatically. This ensures that even during high traffic, your Kling O1 API requests are queued and processed efficiently without you needing to manage GPU scaling.

Transitioning is seamless. Once you have tested your prompts and parameters in the Kling O1 Playground, you can use the exact same inputs (like prompt, start_image_url, and duration) in your code. Simply call the RunComfy HTTP endpoint to integrate Kling O1 directly into your application.

Yes. The Kling O1 Image-to-Video mode is designed for this. You should provide both a start image and an end image in the API parameters. In your prompt, use the references @Image1 and @Image2 to guide Kling O1 on how to transition between these two visual states.