Higgsfield - Result Video

Result documents only get you so far. The advanced and resume nodes hand back an HF_RESULT, which is a bundle of saved file paths - fine as a record, useless as an input. HFResultVideo is the half of the pair that turns one into a native ComfyUI VIDEO object, which means the clip can flow into Save Video, Get Video Components and anything else that accepts the modern VIDEO type.

Why care about the audio half of that? Because the open video models in wide local use are mostly silent - the top comment on one of 2026's best open I2V releases was literally "The king is mute!" - and native audio is one of the few things a paid API route gives you that your local Wan or LTX setup doesn't. If the request generated audio, this node keeps it attached instead of leaving it on the cutting-room floor.

How it works

The generating node has already downloaded the finished file into ComfyUI/output/higgsfield/<your-key-folder>/, so this node does no HTTP and no polling: it opens that file as a native VIDEO object and hands it downstream, audio intact. It also registers a video preview, so your clip plays inside the node. Point it at a request that produced no video and it raises "This request did not produce a video" - that's the signal you wanted Result Images, or that the route returned something the pack can't save (audio-only returns are explicitly out of scope).

One requirement worth stating plainly: this needs a reasonably recent ComfyUI. The VIDEO type and the Save Video node are recent core additions, and the pack uses the current comfy_api.latest interface for them. On an older build you'll get import errors or a missing node type, not a graceful degradation.

Inputs and outputs

One input: result, the HF_RESULT from Custom Model (Advanced) or Resume Request. The preset video node hands you a VIDEO directly and doesn't need this.

Three outputs. video is the VIDEO object - into Save Video to get an MP4 under your output folder with your own naming, or into Get Video Components if you want the frames as IMAGEs and the audio as a separate stream. saved_path is a plain string with the file the pack already wrote, for logging or a downstream script. request_id is the usual handle for Resume Request.

Like the rest of the pack's terminal nodes it doubles as a preview, so a two-node workflow - Advanced Generate into Result Video - is a legitimate way to watch a request without wiring anything else.

Install

Install the pack once:

cd ComfyUI/custom_nodes
git clone https://github.com/w0ver/Higgsfield-api-comfyui-nodes ComfyUI-Higgsfield

Restart and it appears under Higgsfield/Advanced (ComfyUI Manager, search "Higgsfield", does the same thing). Dependencies are a single line - requests - and there are no model files to fetch, since nothing runs locally. Credentials are set with Configure API.bat on Windows or HF_API_KEY_ID / HF_API_KEY_SECRET in the ComfyUI backend's environment elsewhere.

Where people get burned

• The download itself has limits: 1 GiB per file and a ten-minute ceiling. A long, high-resolution Seedance 2.0 clip is the realistic candidate for hitting one. If it happens, the finished request is still in the local database, so queueing Resume Request with the same request_id retries the download without paying again.
• "This request did not produce a video." Usually an image route pointed at the video converter, or a route whose output shape the pack doesn't handle. Check which node you fed it.
• The output folder is load-bearing. This node reads the downloaded file from disk. Move or delete ComfyUI/output/higgsfield/... and the result is gone from the graph's point of view - recoverable with Resume Request, not with a re-queue of the generator (which would bill you again).
• No audio in the output. Check that generate_audio was on for the generation; this node can only preserve what came back.
• Multi-user ComfyUI isn't supported by these nodes at all - they refuse to run rather than share one credential across sessions.