CosyVoice3 — Badini Kurdish TTS (fine-tuned)

A CosyVoice3 model (https://github.com/FunAudioLLM/CosyVoice) fine-tuned for zero-shot text-to-speech in the Badini dialect of Kurdish.

Usage

import sys
sys.path.append("CosyVoice")
sys.path.append("CosyVoice/third_party/Matcha-TTS")
from cosyvoice.cli.cosyvoice import AutoModel
import torchaudio

cosy = AutoModel(model_dir="<path to downloaded repo>")
prompt_wav = "path/to/prompt.wav"
prompt_text = "transcript of the prompt audio"
instruct = "You are a helpful assistant.<|endofprompt|>"

for out in cosy.inference_zero_shot(
    "text to synthesize", instruct + prompt_text, prompt_wav, stream=False
):
    torchaudio.save("output.wav", out["tts_speech"], cosy.sample_rate)