VITS-TH-Model

My first TTS

Finetune colab

Example Code

import torch
from transformers import VitsTokenizer, VitsModel, set_seed
import scipy.io.wavfile

device = "cuda" if torch.cuda.is_available() else "cpu"

model = VitsModel.from_pretrained("meguscx/VITS-TH-Model").to(device)
tokenizer = VitsTokenizer.from_pretrained("meguscx/VITS-TH-Model")

text = "การเรียนรู้ภาษาใหม่ช่วยเปิดโลกทัศน์ให้กว้างขึ้น"

inputs = tokenizer(text=text, return_tensors="pt").to(device)

set_seed(456)

with torch.no_grad():
    outputs = model(**inputs)

waveform = outputs.waveform[0].cpu().numpy()

scipy.io.wavfile.write(
    "test.wav",
    rate=model.config.sampling_rate,
    data=waveform
)

print("Saved successfully.")

this model train 6 hr with only 1094 voice data (single speaker)
so voice maybe not too good and sometimes sound weird or unnatural
because dataset is small ;-;

Sample Audio

<audio controls> <source src="https://huggingface.co/meguscx/VITS-TH-Model/resolve/main/test.wav" type="audio/wav"> Your browser does not support the audio element. </audio>