Kwaivgi Kling Video O3 Pro Text To Video (Channel Edition)

If you're choosing a single node from BizyAirPlus's Kling shelf to do serious text-to-video, this is the one. Kling O3 is the current flagship line - the model people name first when they talk about replacing Sora 2 - and Pro is its top tier. You get the best prompt adherence, the most coherent motion, and native audio, all for the price of one node in your graph and a cloud request behind it.

What it is

A BizyAirPlus "Channel Edition" node: BizyAir routes you to Kling's model, you never touch Kling's own web app or API contract, and the generation happens entirely in the cloud. Kling ships no open weights, so "using it in ComfyUI" only exists because of these reseller-style channel nodes. Everything runs on your BizyAir account - an API key from bizyair.ai, credits to spend, no local GPU involvement.

How it works

Prompt in, video out, with O3 doing the heavy lifting remotely. The interesting part for storytellers is that this is a multi-shot capable model: turn multi_shot on and one request returns a video broken into planned shots instead of a single continuous take. You can let the model plan the breakdown (shot_type = "intelligence") or take the wheel and describe each shot yourself in the multi_prompt JSON (shot_type = "customize"). For anything longer than a few seconds, that's a genuinely different workflow than a plain T2V node - you're basically scripting.

Inputs that matter

• prompt - required, and it's doing double duty: the scene and (if you're in multi-shot mode) the source the model plans shots from. Write action and camera explicitly.
• duration - 1–15 seconds, default 5. Note this node lets you go down to a 1-second clip.
• aspect_ratio - output shape, default 16:9.
• sound - on by default; O3 generates audio with the video.
• multi_shot - on by default. This is the hidden cost lever: multi-shot requests render more and bill more.
• multi_prompt - JSON of per-shot prompts, used when you set shot_type to "customize".

api_config overrides the endpoint/key per request; skip_error outputs a placeholder on failure rather than aborting the queue.

Outputs

video → SaveVideo or preview. request_id is the cloud job ID for support/billing. urls gives direct download links to the rendered MP4s.

Installing

ComfyUI Manager → "BizyAirPlus", or:

cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart, toggle the BizyAirPlus button ON, and enter the API key when prompted. No model files, no VRAM requirements - the pack installs bizyair-cloudberry and bizytrd from PyPI on startup.

Where people get burned

Two things, both about the bill. First, multi_shot defaults to on - great for storytelling, expensive when you just wanted a single clip; flip it off for quick takes. Second, Pro is the premium tier, and premium metered video adds up fast; community threads regularly flag Kling's newer tiers as the priciest video option on the market. Prototype on the Std node, then spend on Pro once you've locked the prompt and seed.