VibeVoice Load Text From File

Every VibeVoice workflow starts the same way: text in, audio out. And the fastest way to make that text input annoying is to hand-type a 2,000-word script into a multiline box, or to paste it and then watch the node's inputs list scroll. "VibeVoice Load Text From File" is the boring fix: pick a .txt file from a dropdown, and it outputs the file's contents as a text string you wire straight into the text input of a speaker node.

It's the smallest node in the pack - one input, one output - and honestly, you can skip it for short one-liners. Where it earns its keep is real projects: a script you keep editing in a proper text editor between runs, a long-form narration that changes daily, or a workflow you want to rerun on different scripts by swapping a file instead of editing a node.

How it works

The node scans ComfyUI's standard directories - input, output, and temp - for .txt files and lists them in the dropdown, prefixed by which directory each lives in (so you get entries like input/script.txt). Pick one and it loads the contents; the output is a plain text string. The file picker also validates your selection, so a stale or moved file gives you a clear error instead of a silent empty run.

One neat ComfyUI detail: when you wire this into the text input of a speaker node, that text input greys out - the node knows it's being fed and won't fight your connection with its default text.

Inputs and output

file - the dropdown of available .txt files (format: directory/filename). "No text files found in any directory" means you haven't dropped one in input, output, or temp yet.
Output text - the file contents, ready for the Single or Multiple Speakers node.

Only .txt is supported, so convert anything else before you drop it in.

A workflow that shows it off

Put a script with speaker labels in ComfyUI/input/:

[1]: Welcome to the show.
[2]: Thanks for having me back.

Then wire: Load Text From File → VibeVoice Multiple Speakers → Save Audio. Change the script file, hit run, get a new conversation - no node editing. That's the whole loop, and it scales to multi-scene narration where editing a .txt in your editor is far friendlier than untangling a giant text box.

Install

It ships in the same pack as the speaker nodes, so there's no separate install: ComfyUI Manager search "VibeVoice", or git clone https://github.com/Enemyx-net/VibeVoice-ComfyUI into ComfyUI/custom_nodes/, restart, and refresh the browser so the dropdown rescans your folders. No extra dependencies beyond the pack's own. And the file-scans-on-refresh behavior means a file you add while ComfyUI is running won't appear until you refresh the page - one of those tiny gotchas that costs people five minutes the first time.