Input FilePath

Every FasterWhisper transcription in this pack starts here, and it's the least glamorous node you'll ever wire up. It's a text box that holds a file path, validates that the file actually exists, and hands the path down the line. That's it. But it exists for a real reason: the transcription node requires its audio input as a FILEPATH type, and this is the clean, typed way to feed it one - rather than converting a plain string widget and hoping the types line up.

How it works. One required input: filepath, a plain text field (not multiline). Type an absolute path to an audio or video file - C:\Your_File_Path\test.mp3 on Windows, /home/you/clips/clip.mp4 elsewhere - and the node checks os.path.exists(). If the file isn't there, it raises a File not found: <path> error and the queue stops, which is honestly the most useful thing it does: it catches a typo at the source instead of letting a confusing error surface later in the transcription node. The output is filepath (type FILEPATH), wired straight into FasterWhisperTranscription's audio input.

The honesty bit. This node is a placeholder and the author knows it. The README's own todo list says the "better version" would have an upload button - you'd pick the file visually instead of pasting a path. Until that lands, you're typing absolute paths by hand. That's the whole UX, so don't go looking for a file browser; there isn't one. Windows users get the extra fun of escaping backslashes in a text field, and the included example workflow shows the pain literally (C:\Your_File_Path\test.mp3).

Where it fits. The pack's example workflow is: InputFilePath → FasterWhisperTranscription → FasterWhisperToSubtitle → SaveSubtitle. This node is deliberately first and it's not an output node - nothing gets written, it just feeds the graph. If you have a more convenient path source from another pack (a folder loader, a "pick latest file" node), you can almost certainly skip this one, since the FILEPATH type is what matters downstream, not this specific producer.

Install & troubleshooting. Install is the pack's standard - ComfyUI Manager search "ComfyUI-faster-whisper", or cd ComfyUI/custom_nodes && git clone https://github.com/jhj0517/ComfyUI-faster-whisper then pip install -r requirements.txt. No model downloads, no dependencies beyond the pack's. The only real failure mode is a wrong path, and the error message tells you exactly what's wrong: the file wasn't found. Double-check that you typed an absolute path (not a relative one), that any Windows backslashes are intact, and that the file isn't still mid-download when you queue. Everything after that is the transcription node's problem.