☁️BizyAir Kling 2.6 Text To Video

What it is

The newer sibling of the 2.1 text-to-video node, and the meaningful upgrade is right there in the input list: sound. Kling 2.6 can generate synced audio alongside the video, not just picture. That puts it in a genuinely small club - native audio-with-video generation is still a rare capability even among closed models, and it's the kind of thing that usually justifies reaching for a paid API over a local open model in the first place. Same deal as the rest of this pack's Kling nodes: it's Kuaishou's model, wrapped so you call it and pay through your existing BizyAir account instead of a separate Kling subscription.

The inputs that matter

• prompt - describe the scene and its motion.
• sound - boolean, off by default. Turn it on to generate audio alongside the video rather than a silent clip. This is the headline feature over the 2.1 node.
• model_name - locked to kling-v2-6.
• duration - 5 or 10 seconds.
• aspect_ratio - 16:9, 9:16, or 1:1.

Outputs: video, and bizyair_model_name as a passthrough label reporting which backend model version actually served the request.

How to install it

ComfyUI Manager: search "BizyAir," install, restart. Manual route:

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

You need a BizyAir API key set up before this node - or any other API node in the pack - will run. There's no local execution path for Kling at all; the entire node exists to be a thin client for the cloud call.

Common issues

Turning on sound and getting silence, or audio that doesn't match the visual. Native audio generation is still a fairly immature capability across the industry - treat it as a genuine but imperfect feature, and don't be surprised if a specific take needs a re-roll to get audio that actually syncs with what's happening on screen. If audio quality matters more than convenience, a dedicated audio pass afterward is still the more reliable route.

Cost adds up fast with sound and longer durations both on. Audio generation on top of video generation is a heavier request than silent video alone on most APIs of this kind - check your account balance before queuing several 10-second, sound-on generations back to back.

Prompt describes the picture but not the sound, and you get generic ambient noise. If you want specific audio content (dialogue, a particular sound effect), describe it in the prompt explicitly the same way you'd describe visual content - don't assume the model will infer appropriate audio from visual description alone.

Job fails outright. Check your API key and credit balance first; a generic failure on any of this pack's API nodes is more often an auth or billing issue on the BizyAir side than a problem with your prompt.