☁️BizyAir Kling O1 Image To Video

What it is

This is Kling's newer O1 tier doing image-to-video, and the feature worth knowing about upfront is that it takes an optional last frame in addition to the required first one. That's keyframe-to-video: give it a starting image and an ending image, and it interpolates the motion between them, rather than just animating outward from a single frame the way most image-to-video tools do. It's a genuinely more controllable way to get a specific shot than text-to-video, since you're constraining both ends of the clip rather than hoping the model lands somewhere close to what you wanted.

As with the rest of this pack's Kling nodes, it's Kuaishou's closed model accessed through your BizyAir account rather than a separate Kling subscription.

The inputs that matter

• prompt - describes the motion and content of the clip.
• first_frame_image - required. Your starting frame.
• last_frame_image - optional. If you provide one, the model interpolates from first to last rather than freely animating outward; leave it empty for open-ended motion from just the first frame.
• duration - 5 or 10 seconds.
• aspect_ratio - 16:9, 9:16, or 1:1.

Outputs: video, and bizyair_model_name, which for this endpoint reports kling-o1 as the resolved backend model.

How to install it

ComfyUI Manager: search "BizyAir," install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAir.git

A BizyAir API key is required - Kling has no local path at all, this node is purely a client for the hosted service.

Common issues

Motion between first and last frame looks unnatural or jumps. Keyframe interpolation works best when the two frames represent a plausible continuous motion - if the subject's pose or the camera position differs too drastically between them, the model has to guess at an in-between that may not read cleanly. Try frames that are closer together in implied time, or lean more on the prompt to describe the intended motion explicitly.

Only using the first frame and skipping last_frame_image entirely. That's a completely valid way to use this node - it just becomes standard single-frame image-to-video at that point, functionally closer to the text-to-video Kling nodes but anchored to your specific starting image instead of generated from scratch.

Aspect ratio mismatch with your source frame(s). Only three ratios are offered here (16:9, 9:16, 1:1) - if your source images are a different shape, expect cropping or padding rather than a stretch; pick the closest match upfront to avoid surprises.

Job fails outright. Check API key and account credits before anything else - that covers the large majority of failures on this pack's API-wrapped nodes, well before a genuinely bad input image or prompt does.