Parler-TTS is a lightweight text-to-speech (TTS) model that can generate high-quality, natural sounding speech in the style of a given speaker (gender, pitch, speaking style, etc). It is a reproduction of work from the paper Natural language guidance of high-fidelity text-to-speech with synthetic annotations by Dan Lyth and Simon King, from Stability AI and Edinburgh University respectively.
Contrarily to other TTS models, Parler-TTS is a fully open-source release. All of the datasets, pre-processing, training code and weights are released publicly under permissive license, enabling the community to build on our work and develop their own powerful TTS models.
This repository contains the inference and training code for Parler-TTS. It is designed to accompany the Data-Speech repository for dataset annotation.
Important
08/08/2024: We are proud to release two new Parler-TTS checkpoints:
- Parler-TTS Mini, an 880M parameter model.
- Parler-TTS Large, a 2.3B parameter model.
These checkpoints have been trained on 45k hours of audiobook data.
In addition, the code is optimized for much faster generation: we've added SDPA and Flash Attention 2 compatibility, as well as the ability to compile the model.
Parler-TTS has light-weight dependencies and can be installed in one line:
pip install git+https://github.com/huggingface/parler-tts.gitApple Silicon users will need to run a follow-up command to make use the nightly PyTorch (2.4) build for bfloat16 support:
pip3 install --pre torch torchaudio --index-url https://download.pytorch.org/whl/nightly/cpuTip
You can directly try it out in an interactive demo here!
Using Parler-TTS is as simple as "bonjour". Simply install the library once:
pip install git+https://github.com/huggingface/parler-tts.gitParler-TTS has been trained to generate speech with features that can be controlled with a simple text prompt, for example:
import torch
from parler_tts import ParlerTTSForConditionalGeneration
from transformers import AutoTokenizer
import soundfile as sf
device = "cuda:0" if torch.cuda.is_available() else "cpu"
model = ParlerTTSForConditionalGeneration.from_pretrained("parler-tts/parler-tts-mini-v1").to(device)
tokenizer = AutoTokenizer.from_pretrained("parler-tts/parler-tts-mini-v1")
prompt = "Hey, how are you doing today?"
description = "A female speaker delivers a slightly expressive and animated speech with a moderate speed and pitch. The recording is of very high quality, with the speaker's voice sounding clear and very close up."
input_ids = tokenizer(description, return_tensors="pt").input_ids.to(device)
prompt_input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
audio_arr = generation.cpu().numpy().squeeze()
sf.write("parler_tts_out.wav", audio_arr, model.config.sampling_rate)To ensure speaker consistency across generations, this checkpoint was also trained on 34 speakers, characterized by name (e.g. Jon, Lea, Gary, Jenna, Mike, Laura).
To take advantage of this, simply adapt your text description to specify which speaker to use: Jon's voice is monotone yet slightly fast in delivery, with a very close recording that almost has no background noise.
import torch
from parler_tts import ParlerTTSForConditionalGeneration
from transformers import AutoTokenizer
import soundfile as sf
device = "cuda:0" if torch.cuda.is_available() else "cpu"
model = ParlerTTSForConditionalGeneration.from_pretrained("parler-tts/parler-tts-mini-v1").to(device)
tokenizer = AutoTokenizer.from_pretrained("parler-tts/parler-tts-mini-v1")
prompt = "Hey, how are you doing today?"
description = "Jon's voice is monotone yet slightly fast in delivery, with a very close recording that almost has no background noise."
input_ids = tokenizer(description, return_tensors="pt").input_ids.to(device)
prompt_input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
audio_arr = generation.cpu().numpy().squeeze()
sf.write("parler_tts_out.wav", audio_arr, model.config.sampling_rate)Tips:
- Include the term "very clear audio" to generate the highest quality audio, and "very noisy audio" for high levels of background noise
- Punctuation can be used to control the prosody of the generations, e.g. use commas to add small breaks in speech
- The remaining speech features (gender, speaking rate, pitch and reverberation) can be controlled directly through the prompt
We've set up an inference guide to make generation faster. Think SDPA, torch.compile and streaming!
Yoach.Lacombe.s.Video.-.Apr.10.2024.1.mp4
The training folder contains all the information to train or fine-tune your own Parler-TTS model. It consists of:
- 1. An introduction to the Parler-TTS architecture
- 2. The first steps to get started
- 3. A training guide
Important
TL;DR: After having followed the installation steps, you can reproduce the Parler-TTS Mini v1 training recipe with the following command line:
accelerate launch ./training/run_parler_tts_training.py ./helpers/training_configs/starting_point_v1.jsonImportant
You can also follow this fine-tuning guide on a mono-speaker dataset example.
This library builds on top of a number of open-source giants, to whom we'd like to extend our warmest thanks for providing these tools!
Special thanks to:
- Dan Lyth and Simon King, from Stability AI and Edinburgh University respectively, for publishing such a promising and clear research paper: Natural language guidance of high-fidelity text-to-speech with synthetic annotations.
- the many libraries used, namely 🤗 datasets, 🤗 accelerate, jiwer, wandb, and 🤗 transformers.
- Descript for the DAC codec model
- Hugging Face 🤗 for providing compute resources and time to explore!
If you found this repository useful, please consider citing this work and also the original Stability AI paper:
@misc{lacombe-etal-2024-parler-tts,
author = {Yoach Lacombe and Vaibhav Srivastav and Sanchit Gandhi},
title = {Parler-TTS},
year = {2024},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/huggingface/parler-tts}}
}
@misc{lyth2024natural,
title={Natural language guidance of high-fidelity text-to-speech with synthetic annotations},
author={Dan Lyth and Simon King},
year={2024},
eprint={2402.01912},
archivePrefix={arXiv},
primaryClass={cs.SD}
}
Contributions are welcome, as the project offers many possibilities for improvement and exploration.
Namely, we're looking at ways to improve both quality and speed:
- Datasets:
- Train on more data
- Add more features such as accents
- Training:
- Add PEFT compatibility to do Lora fine-tuning.
- Add possibility to train without description column.
- Add notebook training.
- Explore multilingual training.
- Explore mono-speaker finetuning.
- Explore more architectures.
- Optimization:
- Compilation and static cache
- Support to FA2 and SDPA
- Evaluation:
- Add more evaluation metrics
