RH Audio Uploader

Audio-driven workflows are having a moment - lip-sync, talking-avatar, voice-to-video pipelines - and a lot of them are heavy enough that you'd rather run them on rented GPUs than your own. RH_AudioUploader is the piece that lets you do that on RunningHub: it takes a local audio file and uploads it, returning the remote filename the cloud workflow will reference.

It's part of ComfyUI_RH_APICall, the plugin that runs workflows on the RunningHub cloud and returns results into your local ComfyUI. And it exists for the same structural reason as the image and video uploaders: the workflow executes on RunningHub's servers, which have no access to your disk. A community thread put the general problem plainly - local file nodes "can't connect to your PC" once you're running in the cloud. So audio your remote graph needs has to be pushed up first, and this is the node that pushes it.

How it works, and why it takes a path

The input is a path string, not an audio wire. Images in this pack travel as the normal ComfyUI IMAGE type, but audio (like video) comes in as a filesystem path - you tell the node where the file lives and it reads and uploads it. Point it at something like C:\audio\voice.wav, and it sends that file to RunningHub through the API and hands back the name it now has on their side. That filename is what you plug into an override so a specific node in the cloud workflow uses your clip.

The inputs and output that matter

• audio (STRING) - the path to the local audio file to upload. A plain string; type or paste the full path to the file.
• apiConfig (STRUCT) - the credential bundle from RH_SettingsNode, so the uploader knows which account and endpoint to target. Wire your Settings node in here; without it, nothing uploads.

The output is filename (STRING) - the remote name of your uploaded audio. Route it through a NodeInfoList override into the main execute node, which runs the cloud workflow and consumes it.

Installing it

One-shot install for the whole pack, and it's light because the compute is remote:

• ComfyUI Manager: search ComfyUI_RH_APICall, install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/HM-RunningHub/ComfyUI_RH_APICall, then restart ComfyUI.

Set your API key and workflow ID on the RH_SettingsNode (this node just consumes that config), and check the pack's examples/ folder for a graph that wires an upload into a full run.

Where people get stuck

The usual suspects are a bad path - a typo, a relative path where an absolute one was needed, or Windows backslash quirks - and credentials, which fail at the first networked node: a blank apiKey, or the .cn/.ai mismatch where your account and your base_url disagree on domain. Confirm the path points at a real file, and confirm the Settings node matches the domain you registered on.

If the upload goes through but the audio doesn't seem to affect the result, that's the override, not the uploader - your NodeInfoList entry is targeting the wrong node ID or field in the cloud workflow, so it ran on a default. Fix the override to point at the actual audio-loading node as it exists on RunningHub. And the standing note for the pack: your file leaves your machine for a third party's servers, so upload only what you're happy handing over.