Kling VIDEO 3.0 Standard

Kling 3.0 Standard is what you queue when you're still figuring out what you want. Same family as 3.0 Pro, same interface - prompt, dimensions, duration - but it's the cheaper tier, which makes it the natural place to burn through prompt ideas before you spend real money on a Pro render. In the local world it's the difference between testing a LoRA at 20 steps and committing to a full 40-step batch.

It's a closed model served through Runware's cloud, so no weights, no VRAM, no nothing. Your machine just holds the workflow while Kuaishou's servers do the math, and the node reports back with a native VIDEO plus a printed cost per run. That cost line is your friend here - Standard exists so you can afford to be wrong a few times.

What it shares with Pro

Everything structurally, minus the price. The node sends klingai:kling-video@3-standard with a videoInference task to Runware's REST API via the SDK, then downloads the MP4 and wraps it for ComfyUI. The widgets you'll actually set:

• positivePrompt and negativePrompt - both present; the negative is ignored when guidance is disabled (CFG ≤ 1).
• width / height - direct pixel widgets, 1024 default.
• duration - a plain 3–15 second slider.
• characterOrientation - pull orientation from image or video reference if you're feeding one.
• sound + keepOriginalSound - generate audio or carry the reference's.
• numberResults - 1–4 variations per queue. Set to 1 while testing; you can always re-queue.

Conditioning sockets - referenceImages, frameImages, referenceVideos, and video (URL/UUID) - work the same as Pro, so a workflow you build against Standard transfers up to Pro without rewiring. That's the argument for starting here.

The output

A single video socket (type VIDEO) into Preview Video or Save Video. Older ComfyUI builds degrade it to a file path string - update ComfyUI and it becomes the native video type.

Installing it

One-time pack install, shared by every node:

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

Restart, set the key in ComfyUI Settings → Runware API key (or RUNWARE_API_KEY, or runware auth login), and you're live. Nothing heavy installs; no model files to fetch.

Where people get burned

Mostly expectations. Standard output is fine but it's not Pro - don't judge the model by its cheapest tier and write the whole thing off, and don't ship a Standard render when the client asked for the good one. The video input still wants a URL or UUID, never a local path. And the "No Runware API key" error means the Settings field didn't save, not that the pack is broken.