Timbre音频加载器

Every node in this pack that clones a voice needs an AUDIO-typed reference clip wired in. If you're testing a bunch of voices, re-loading a file with a generic Load Audio node every time gets old fast. Timbre Audio Loader is the pack's answer: drop your reference clips into one folder (ComfyUI-Index-TTS/TimbreModel), and they show up as a simple dropdown you pick from instead of browsing your filesystem each run.

It ships with a grab bag of sample voices already in that folder for you to try - a mix of Chinese internet-culture reference clips (a comedian, a singer, various character-style voices pulled from games and novels) alongside a "no audio file" placeholder option. They're there to get you generating something on your first run before you've recorded or sourced your own reference audio.

Inputs and outputs that matter

Required: audio_file - the dropdown, populated by scanning the TimbreModel folder. It lists whatever's in there: .wav, .mp3, and .flac are all fair game, based on the bundled sample set. refresh (BOOLEAN, default off) - flip this if you've just dropped a new file into the folder and it isn't showing up in the list yet; like most ComfyUI loader nodes, the dropdown is built from a directory scan that doesn't automatically re-run mid-session.

Output: a single AUDIO - wire it directly into reference_audio on Index TTS / Index TTS 2 nodes, or narrator_audio / characterN_audio on the Pro nodes, exactly like any other audio source.

Install

ComfyUI Manager: search ComfyUI-Index-TTS, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/chenpipi0807/ComfyUI-Index-TTS.git
cd ComfyUI-Index-TTS
pip install -r requirements.txt

No model download needed for this node specifically - it's just a file loader. To use your own voices instead of (or alongside) the bundled samples, drop .wav/.mp3/.flac files straight into ComfyUI/custom_nodes/ComfyUI-Index-TTS/TimbreModel/ and hit refresh on the node (or restart ComfyUI) to see them appear in the dropdown.

Where people get burned

The whole failure mode of this node is "I added a file and it's not there" - that's the refresh-cache issue above, not a broken install; toggle refresh or restart ComfyUI and it'll pick up new files. Beyond that, remember this node only loads the clip - it doesn't tell you anything about its quality. A quiet, noisy, or heavily compressed reference file loaded through here will clone just as poorly as one loaded any other way; if your generated voice sounds off, run the source clip through Audio Cleaner or re-record it before assuming a TTS node is at fault. And if you're sharing a workflow with someone else, know that this node's dropdown only works if the same files exist in their TimbreModel folder too - the audio isn't embedded in the workflow JSON, just the filename reference, so a shared workflow using a bundled sample voice will work out of the box, but one using your own custom drop-in won't until you send the file along with it.