by atul10
Open source · 250 downloads · 1 likes
The *Nepali Male v1* model is a text-to-speech (TTS) system based on the VITS architecture, specifically designed to generate highly expressive male Nepali voices. It converts text into natural-sounding speech by integrating a text encoder, a stochastic duration predictor, and a spectrogram decoder, all optimized through adversarial learning techniques to achieve realistic audio quality. Thanks to its end-to-end approach, it captures the nuances of prosody and rhythm, enabling natural variations for the same input text. Ideal for applications such as voice assistants, audiobooks, or accessibility tools, it stands out for its ability to produce varied intonations while remaining faithful to the semantic content. Its multilingual training (Nepali and Hindi) enhances its versatility, while its non-deterministic nature offers creative flexibility to adapt the voice to different contexts.
Nepali language
VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is an end-to-end speech synthesis model that predicts a speech waveform conditional on an input text sequence. It is a conditional variational autoencoder (VAE) comprised of a posterior encoder, decoder, and conditional prior.
A set of spectrogram-based acoustic features are predicted by the flow-based module, which is formed of a Transformer-based text encoder and multiple coupling layers. The spectrogram is decoded using a stack of transposed convolutional layers, much in the same style as the HiFi-GAN vocoder. Motivated by the one-to-many nature of the TTS problem, where the same text input can be spoken in multiple ways, the model also includes a stochastic duration predictor, which allows the model to synthesise speech with different rhythms from the same input text.
The model is trained end-to-end with a combination of losses derived from variational lower bound and adversarial training. To improve the expressiveness of the model, normalizing flows are applied to the conditional prior distribution. During inference, the text encodings are up-sampled based on the duration prediction module, and then mapped into the waveform using a cascade of the flow module and HiFi-GAN decoder. Due to the stochastic nature of the duration predictor, the model is non-deterministic, and thus requires a fixed seed to generate the same speech waveform.
TTS is available in the 🤗 Transformers library from version 4.33 onwards. To use this checkpoint, first install the latest version of the library:
pip install --upgrade transformers accelerate
Then, run inference with the following code-snippet:
from transformers import VitsModel, AutoTokenizer
import torch
model = VitsModel.from_pretrained("procit001/nepali_male_v1")
tokenizer = AutoTokenizer.from_pretrained("procit001/nepali_male_v1")
text = "म पनि जान्छु है त अहिले लाई"
inputs = tokenizer(text, return_tensors="pt")
with torch.no_grad():
output = model(**inputs).waveform
The resulting waveform can be saved as a .wav file:
import scipy
scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=output)
Or displayed in a Jupyter Notebook / Google Colab:
from IPython.display import Audio
Audio(output, rate=model.config.sampling_rate)
The model is licensed as procitBV v1.
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.