
* Fixed typo: insted to instead * Fixed typo: relase to release * Fixed typo: nighlty to nightly * Fixed typos: versatible, benchamarks, becnhmark to versatile, benchmark, benchmarks * Fixed typo in comment: quantizd to quantized * Fixed typo: architecutre to architecture * Fixed typo: contibution to contribution * Fixed typo: Presequities to Prerequisites * Fixed typo: faste to faster * Fixed typo: extendeding to extending * Fixed typo: segmetantion_maps to segmentation_maps * Fixed typo: Alternativelly to Alternatively * Fixed incorrectly defined variable: output to output_disabled * Fixed typo in library name: tranformers.onnx to transformers.onnx * Fixed missing import: import tensorflow as tf * Fixed incorrectly defined variable: token_tensor to tokens_tensor * Fixed missing import: import torch * Fixed incorrectly defined variable and typo: uromaize to uromanize * Fixed incorrectly defined variable and typo: uromaize to uromanize * Fixed typo in function args: numpy.ndarry to numpy.ndarray * Fixed Inconsistent Library Name: Torchscript to TorchScript * Fixed Inconsistent Class Name: OneformerProcessor to OneFormerProcessor * Fixed Inconsistent Class Named Typo: TFLNetForMultipleChoice to TFXLNetForMultipleChoice * Fixed Inconsistent Library Name Typo: Pytorch to PyTorch * Fixed Inconsistent Function Name Typo: captureWarning to captureWarnings * Fixed Inconsistent Library Name Typo: Pytorch to PyTorch * Fixed Inconsistent Class Name Typo: TrainingArgument to TrainingArguments * Fixed Inconsistent Model Name Typo: Swin2R to Swin2SR * Fixed Inconsistent Model Name Typo: EART to BERT * Fixed Inconsistent Library Name Typo: TensorFLow to TensorFlow * Fixed Broken Link for Speech Emotion Classification with Wav2Vec2 * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed minor missing word Typo * Fixed Punctuation: Two commas * Fixed Punctuation: No Space between XLM-R and is * Fixed Punctuation: No Space between [~accelerate.Accelerator.backward] and method * Added backticks to display model.fit() in codeblock * Added backticks to display openai-community/gpt2 in codeblock * Fixed Minor Typo: will to with * Fixed Minor Typo: is to are * Fixed Minor Typo: in to on * Fixed Minor Typo: inhibits to exhibits * Fixed Minor Typo: they need to it needs * Fixed Minor Typo: cast the load the checkpoints To load the checkpoints * Fixed Inconsistent Class Name Typo: TFCamembertForCasualLM to TFCamembertForCausalLM * Fixed typo in attribute name: outputs.last_hidden_states to outputs.last_hidden_state * Added missing verbosity level: fatal * Fixed Minor Typo: take To takes * Fixed Minor Typo: heuristic To heuristics * Fixed Minor Typo: setting To settings * Fixed Minor Typo: Content To Contents * Fixed Minor Typo: millions To million * Fixed Minor Typo: difference To differences * Fixed Minor Typo: while extract To which extracts * Fixed Minor Typo: Hereby To Here * Fixed Minor Typo: addition To additional * Fixed Minor Typo: supports To supported * Fixed Minor Typo: so that benchmark results TO as a consequence, benchmark * Fixed Minor Typo: a To an * Fixed Minor Typo: a To an * Fixed Minor Typo: Chain-of-though To Chain-of-thought
8.1 KiB
VITS
Overview
The VITS model was proposed in Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech by Jaehyeon Kim, Jungil Kong, Juhee Son.
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.
The abstract from the paper is the following:
Several recent end-to-end text-to-speech (TTS) models enabling single-stage training and parallel sampling have been proposed, but their sample quality does not match that of two-stage TTS systems. In this work, we present a parallel end-to-end TTS method that generates more natural sounding audio than current two-stage models. Our method adopts variational inference augmented with normalizing flows and an adversarial training process, which improves the expressive power of generative modeling. We also propose a stochastic duration predictor to synthesize speech with diverse rhythms from input text. With the uncertainty modeling over latent variables and the stochastic duration predictor, our method expresses the natural one-to-many relationship in which a text input can be spoken in multiple ways with different pitches and rhythms. A subjective human evaluation (mean opinion score, or MOS) on the LJ Speech, a single speaker dataset, shows that our method outperforms the best publicly available TTS systems and achieves a MOS comparable to ground truth.
This model can also be used with TTS checkpoints from Massively Multilingual Speech (MMS) as these checkpoints use the same architecture and a slightly modified tokenizer.
This model was contributed by Matthijs and sanchit-gandhi. The original code can be found here.
Usage examples
Both the VITS and MMS-TTS checkpoints can be used with the same API. Since the flow-based model is non-deterministic, it is good practice to set a seed to ensure reproducibility of the outputs. For languages with a Roman alphabet, such as English or French, the tokenizer can be used directly to pre-process the text inputs. The following code example runs a forward pass using the MMS-TTS English checkpoint:
import torch
from transformers import VitsTokenizer, VitsModel, set_seed
tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
model = VitsModel.from_pretrained("facebook/mms-tts-eng")
inputs = tokenizer(text="Hello - my dog is cute", return_tensors="pt")
set_seed(555) # make deterministic
with torch.no_grad():
outputs = model(**inputs)
waveform = outputs.waveform[0]
The resulting waveform can be saved as a .wav
file:
import scipy
scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=waveform)
Or displayed in a Jupyter Notebook / Google Colab:
from IPython.display import Audio
Audio(waveform, rate=model.config.sampling_rate)
For certain languages with a non-Roman alphabet, such as Arabic, Mandarin or Hindi, the uroman
perl package is required to pre-process the text inputs to the Roman alphabet.
You can check whether you require the uroman
package for your language by inspecting the is_uroman
attribute of
the pre-trained tokenizer
:
from transformers import VitsTokenizer
tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
print(tokenizer.is_uroman)
If the is_uroman attribute is True
, the tokenizer will automatically apply the uroman
package to your text inputs, but you need to install uroman if not already installed using:
pip install --upgrade uroman
Note: Python version required to use uroman
as python package should be >= 3.10
.
You can use the tokenizer as usual without any additional preprocessing steps:
import torch
from transformers import VitsTokenizer, VitsModel, set_seed
import os
import subprocess
tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-kor")
model = VitsModel.from_pretrained("facebook/mms-tts-kor")
text = "이봐 무슨 일이야"
inputs = tokenizer(text=text, return_tensors="pt")
set_seed(555) # make deterministic
with torch.no_grad():
outputs = model(inputs["input_ids"])
waveform = outputs.waveform[0]
If you don't want to upgrade to python >= 3.10
, then you can use the uroman
perl package to pre-process the text inputs to the Roman alphabet.
To do this, first clone the uroman repository to your local machine and set the bash variable UROMAN
to the local path:
git clone https://github.com/isi-nlp/uroman.git
cd uroman
export UROMAN=$(pwd)
You can then pre-process the text input using the following code snippet. You can either rely on using the bash variable
UROMAN
to point to the uroman repository, or you can pass the uroman directory as an argument to the uromanize
function:
import torch
from transformers import VitsTokenizer, VitsModel, set_seed
import os
import subprocess
tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-kor")
model = VitsModel.from_pretrained("facebook/mms-tts-kor")
def uromanize(input_string, uroman_path):
"""Convert non-Roman strings to Roman using the `uroman` perl package."""
script_path = os.path.join(uroman_path, "bin", "uroman.pl")
command = ["perl", script_path]
process = subprocess.Popen(command, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
# Execute the perl command
stdout, stderr = process.communicate(input=input_string.encode())
if process.returncode != 0:
raise ValueError(f"Error {process.returncode}: {stderr.decode()}")
# Return the output as a string and skip the new-line character at the end
return stdout.decode()[:-1]
text = "이봐 무슨 일이야"
uromanized_text = uromanize(text, uroman_path=os.environ["UROMAN"])
inputs = tokenizer(text=uromanized_text, return_tensors="pt")
set_seed(555) # make deterministic
with torch.no_grad():
outputs = model(inputs["input_ids"])
waveform = outputs.waveform[0]
VitsConfig
autodoc VitsConfig
VitsTokenizer
autodoc VitsTokenizer - call - save_vocabulary
VitsModel
autodoc VitsModel - forward