XTTS is receiving input, but throwing error "no wav files found in speakers"

2024-09-09 18:52:54.294 | ERROR | xtts_api_server.server:tts_to_audio:317 - no wav files found in speakers/ INFO: 127.0.0.1:44580 - "POST /tts_to_audio/ HTTP/1.1" 500 Internal Server Error

arch linux.

I installed using pip, which now seems like a bad idea because i don't know what got installed where. i just know it downloaded my 100th copy of pytorch somewhere on my hard drive and a pickletensor.

from the documentation:

How to add speaker

By default the speakers folder should appear in the folder, you need to put there the wav file with the voice sample, you can also create a folder and put there several voice samples, this will give more accurate results Selecting Folder You can change the folders for speakers and the folder for output via the API.

What folder? I'm running this from a command line, installed via pip so don't know where pip put any such folders. I figured since it installed pytorch and a model pickletensor that this program generates TTS.

If I have kobold and i want text to speech, should I be using xtts-api-server or alltalk or something else?

I am using the python (pip) setup, like this (for arch):

sudo pacman -Syu python python-virtualenv portaudio
python -m venv venv
source venv/bin/activate
pip install xtts-api-server
pip install torch==2.5.1+cu118 torchaudio==2.5.1+cu118 --index-url https://download.pytorch.org/whl/cu118

Then you need to create a speakers folder (within the same directory where you've executed the previous calls mkdir speakers

Than you need to download and store some example wav files into this folder, have a look at the examples: https://github.com/daswer123/xtts-api-server/tree/main/example

This works for me now - I had the same error previously, because pip install ... doesn't comes with example wave files :-).