☁️BizyAir Kling 2.1 Text To Video

What it is

Kling is Kuaishou's closed commercial video model - one of the names that comes up constantly whenever people compare local video generation (Wan, LTX) against what the paid services can do, and it's routinely rated near the top of that comparison. It's not open, there's no local install, and getting at it normally means signing up for a separate Kling account and API key. This node skips that: it's a BizyAir wrapper around Kling 2.1's text-to-video endpoint, so you call it like any other node and pay for it out of your existing BizyAir credit balance instead of managing a second vendor relationship.

The practical value is less about the model itself and more about the workflow: you can generate a scene locally or with another BizyAir node, then hand it off into a Kling call, all inside one ComfyUI graph, instead of bouncing between separate tools and separate billing.

The inputs that matter

• prompt - describe the scene and motion; Kling reads more like a video-generation instruction than a static image prompt, so it's worth describing what happens over the clip, not just what's in frame.
• negative_prompt - optional, what to steer away from.
• model_name - locked to kling-v2-1-master, the specific tier this node targets.
• duration - 5 or 10 seconds. Longer costs proportionally more of your credit balance.
• aspect_ratio - 16:9, 9:16, or 1:1.

You get two outputs: video, and bizyair_model_name, a passthrough label reporting the resolved backend model id BizyAir routed your request to - useful mainly for tracking which underlying model actually served a given generation, since API providers sometimes silently route requests across model versions.

How to install it

ComfyUI Manager: search "BizyAir," install, restart. Or clone it directly:

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

A BizyAir account and API key are non-negotiable here - this isn't optional infrastructure the way it might be for a locally-runnable node, since Kling itself is entirely closed and there's no local path at all.

Common issues

Cost surprise. Closed video APIs in this general class have a reputation for being expensive per clip compared to running an open model yourself - worth checking your account balance and BizyAir's current per-second pricing before queuing up a batch of 10-second generations, rather than after.

Generation doesn't match the prompt's intended motion. Text-to-video models in general are more literal about what's in the shot than how it moves; if the motion looks generic, try describing the camera movement and subject action as explicitly as the subject itself, rather than assuming action verbs alone will carry it.

Job fails with no clear reason. Check API key and credits first - a closed third-party API wrapped through BizyAir has two possible failure points (your BizyAir auth, and Kling's own service on the other end), and a generic failure message is more often the former than the latter.

Wanting image-to-video instead. This specific node is text-only - there's a separate Kling O1 image-to-video node elsewhere in this pack's External APIs section if you need to animate a starting frame instead of generating from scratch.