Get 4K Video

Veo's 4K pass is a distinct, slow job. KIE's own example response for this endpoint says the quiet part out loud - 4k is processing. It should be ready in 5-10 minutes. Please check back shortly. - and it returns a task ID, result URLs and companion image URLs when it's done. This node is the fetch step for that.

There's one wrinkle, and it's worth understanding before you blame the pack: the card's request fields are generic.

Why the card looks empty

Every node in this pack is generated from KIE's API documentation. The pack reads each model's request example and parameter metadata and builds widgets from that. Some KIE pages publish only response examples - this 4K page is one of them - so there's no request schema to read, and the pack falls back to a generic card: a text field, a timeout, and an expert override. KIE Next's own tooltip on the override says exactly that: it's for "APIs whose documentation has no request schema yet."

So the fields you get are:

prompt (required) - a generic string field. On a schema-less card like this, it's the prompt slot, not a description of what the endpoint wants.

expert_override_json - a raw JSON object merged into the request payload. Here this is the working interface: the task identifier and any other documented request field go in as JSON, spelled the way KIE's page for this endpoint spells them.

timeout_seconds - 1200 default, 30–7200.

callback_url - leave empty for normal ComfyUI use.

If a run fails asking for a field the card doesn't show you, the override is where it goes. And if you don't want to hand-write JSON, the same folder's Get 1080P Video card is a properly schematised sibling - a good place to look for the shape of the identifier you need.

Outputs

video - a ComfyUI VIDEO object, usable by Save Video, a preview, or a downstream stage. url - the primary result URL. all_urls_json - every URL reported; KIE's success example for this endpoint includes both result URLs and image URLs, so keeping the full list matters more here than on most nodes. task_id - the job handle. raw_json - the entire response, which is where you confirm whether the pass finished or is still processing. credits_consumed and credits_left - floats, plus the credits receipt the node draws on itself after a successful run.

Install

ComfyUI Manager (search the pack title), or:

cd ComfyUI/custom_nodes
git clone https://github.com/felipederosilva/ComfyUI-KIE-Nodes-Next
pip install -r ComfyUI-KIE-Nodes-Next/requirements.txt

Restart. Requirements: requests, Pillow, numpy, PyYAML. Nothing local, no GPU, metered per call.

Key: Settings → KIE.ai Nodes Next → Connection → Paste / replace KIE.ai API key - validated before it replaces a saved key, masked fingerprint and CONNECTED after. Headless installs should set KIE_API_KEY, which takes priority.

Gotchas

Don't fire it and stare. Five to ten minutes is the provider's own estimate, and the node polls up to timeout_seconds. Queue the fetch as a step you come back to rather than the last thing in a run you're watching.

"Processing" is not an error. The documented response for the in-progress case carries a code and a message, not a failure. Check raw_json to distinguish "wait" from "broken" - and note that a node finishing with empty video and a processing body is behaving, not crashing.

Don't re-submit a 4K pass you can't afford to lose twice. If a fetch times out, the job usually continues on KIE's side; re-running the fetch with the same identifier costs another call, so prefer re-running this node over regenerating the video.

The card can grow widgets. The pack re-reads KIE's documentation about once a day; if the provider ever publishes a request schema for this endpoint, the card tends to gain proper fields after a ComfyUI restart.