Generate Aleph Video

Aleph is Runway's video-editing model - the one their 2026 product line lists alongside Gen-4.5 and their world models. So this node is not a generate-from-nothing card: a source video is a required input, and the prompt describes how that footage should change. If you were looking for text-to-video from Runway, that's the Generate AI Video card; this is the other half.

What that means in a workflow

Video-to-video is the part of closed-model video that has no real local equivalent yet. You bring footage - your own render, a filmed clip, something a local model produced - and the API restyles, relights or edits it while keeping the motion. That makes this the node you reach for when the motion is already right and the look is wrong, the inverse of the usual AI-video problem.

It's Runway on KIE, so the two-part caveat from the KB applies: your source footage and prompt leave your machine, and the provider's filters decide what's editable. The pack hides the URL plumbing, not the network call.

Inputs

videoUrl (required) - a native VIDEO socket. This is the whole premise of the node: connect your source clip. The pack uploads connected media to KIE for you and passes the resulting URL to the API.

prompt (required) - the edit instruction. KIE's docs frame it as how to animate or modify the reference footage, so describe the transformation, not the scene from scratch.

waterMark - a string defaulting to kie.ai. Empty string means none. Set it deliberately.

aspectRatio - 16:9, 9:16, 4:3, 3:4, 1:1, 21:9. Wide options; only you know your delivery format.

seed - INT, default 123456. Documented as the reproducibility control, and the default is a fixed number rather than a random one, so identical runs are at least trying to be identical.

referenceImage - a URL string for a style/content reference image. Worth flagging: unlike most of this pack, this field stayed a plain text URL rather than becoming an image socket, matching KIE's documentation for the endpoint. Public URL only - no local file gets uploaded here.

uploadCn - a boolean for the upload route: false (the default) sends the file to the overseas R2 storage, true sends it to Alibaba Cloud OSS. This is a latency question for anyone outside one of those regions.

timeout_seconds - 1200 default, 30–7200. callback_url - leave empty for normal ComfyUI use.

Outputs

video - a ComfyUI VIDEO object, so it wires into Save Video, a preview, or a second pass. url - the primary result. all_urls_json - all reported URLs, kept because edit jobs can return more than one artifact. task_id - the job handle for follow-up queries. raw_json - the full provider response. credits_consumed and credits_left - floats, plus the credits receipt drawn on the node after a successful run.

Install

ComfyUI Manager can find the pack by 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 to load or run; you pay per call on KIE's side.

Key once: Settings → KIE.ai Nodes Next → Connection → Paste / replace KIE.ai API key. It's verified against KIE before it replaces a saved key, then shows a masked fingerprint and CONNECTED. For servers, KIE_API_KEY in the environment overrides the saved key.

Gotchas

Long footage means long waits. Upload plus processing on a clip is slower than a still-image job; the default 1200-second timeout is generous but not unlimited, so raise it for long sources rather than re-queueing.

Everything costs, including failures that got far enough to process. Do your first run on a short clip.

The referenceImage field wants a hosted URL. A local file pasted into that box won't work - that's the documented field, not a socket.

Menu follows the catalogue. The pack builds its node library from KIE's live API docs on startup and re-checks about daily, so this card's fields can change after a ComfyUI restart. On a reseller, availability moves too.