An 16kHz implementation of HiFi-GAN for soft-vc.

Last update: Dec 27, 2022

Overview

HiFi-GAN

An 16kHz implementation of HiFi-GAN for soft-vc.

Relevant links:

Example Usage

import torch
import numpy as np

# Load checkpoint
hifigan = torch.hub.load("bshall/hifigan:main", "hifigan_hubert_soft").cuda()
# Load mel-spectrogram
mel = torch.from_numpy(np.load("path/to/mel")).unsqueeze(0).cuda()
# Generate
wav, sr = hifigan.generate(mel)

Train

Step 1: Download and extract the LJ-Speech dataset

Step 2: Resample the audio to 16kHz:

usage: resample.py [-h] [--sample-rate SAMPLE_RATE] in-dir out-dir

Resample an audio dataset.

positional arguments:
  in-dir                path to the dataset directory
  out-dir               path to the output directory

optional arguments:
  -h, --help            show this help message and exit
  --sample-rate SAMPLE_RATE
                        target sample rate (default 16kHz)

Step 3: Download the dataset splits and move them into the root of the dataset directory. After steps 2 and 3 your dataset directory should look like this:

LJSpeech-1.1
│   test.txt
│   train.txt
│   validation.txt
├───mels
└───wavs

Note: the mels directory is optional. If you want to fine-tune HiFi-GAN the mels directory should contain ground-truth aligned spectrograms from an acoustic model.

Step 4: Train HiFi-GAN:

usage: train.py [-h] [--resume RESUME] [--finetune] dataset-dir checkpoint-dir

Train or finetune HiFi-GAN.

positional arguments:
  dataset-dir      path to the preprocessed data directory
  checkpoint-dir   path to the checkpoint directory

optional arguments:
  -h, --help       show this help message and exit
  --resume RESUME  path to the checkpoint to resume from
  --finetune       whether to finetune (note that a resume path must be given)

Generate

To generate using the trained HiFi-GAN models, see Example Usage or use the generate.py script:

usage: generate.py [-h] [--model-name {hifigan,hifigan-hubert-soft,hifigan-hubert-discrete}] in-dir out-dir

Generate audio for a directory of mel-spectrogams using HiFi-GAN.

positional arguments:
  in-dir                path to directory containing the mel-spectrograms
  out-dir               path to output directory

optional arguments:
  -h, --help            show this help message and exit
  --model-name {hifigan,hifigan-hubert-soft,hifigan-hubert-discrete}
                        available models

Acknowledgements

This repo is based heavily on https://github.com/jik876/hifi-gan.

You might also like...

Fast Soft Color Segmentation

3 Oct 29, 2022

Permute Me Softly: Learning Soft Permutations for Graph Representations

7 Jul 10, 2022

Multi-task Multi-agent Soft Actor Critic for SMAC

Multi-task Multi-agent Soft Actor Critic for SMAC Overview The CARE formulti-task: Multi-Task Reinforcement Learning with Context-based Representation

8 Sep 30, 2022

[ICLR 2022] Contact Points Discovery for Soft-Body Manipulations with Differentiable Physics

CPDeform Code and data for paper Contact Points Discovery for Soft-Body Manipulations with Differentiable Physics at ICLR 2022 (Spotlight). @InProceed

29 Nov 29, 2022

Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two

512x512 flowers after 12 hours of training, 1 gpu 256x256 flowers after 12 hours of training, 1 gpu Pizza 'Lightweight' GAN Implementation of 'lightwe

1.5k Jan 2, 2023

Implementation of TransGanFormer, an all-attention GAN that combines the finding from the recent GanFormer and TransGan paper

TransGanFormer (wip) Implementation of TransGanFormer, an all-attention GAN that combines the finding from the recent GansFormer and TransGan paper. I

146 Dec 6, 2022

PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement.

DECOR-GAN PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement, Zhiqin Chen, Vladimir G. Kim, Matthew Fish

72 Dec 31, 2022

This is a pytorch implementation of the NeurIPS paper GAN Memory with No Forgetting.

GAN Memory for Lifelong learning This is a pytorch implementation of the NeurIPS paper GAN Memory with No Forgetting. Please consider citing our paper

43 Dec 27, 2022

[CVPR 2021] Pytorch implementation of Hijack-GAN: Unintended-Use of Pretrained, Black-Box GANs

Hijack-GAN: Unintended-Use of Pretrained, Black-Box GANs In this work, we propose a framework HijackGAN, which enables non-linear latent space travers

46 Sep 5, 2022

Comments

is pretrained weight of discriminator of base model available?

Thanks for nice work. @bshall

I'm trying to train hifigan now, but it takes so long training it from scratch using other dataset.

If discriminator of base model is also available, I could start finetuning based on that vocoder. it seems that you released only generator. Could you also release discriminator weights?

opened by seastar105 3
NaN during training when using own dataset
While fine-tuning works as expected, doing regular training with a dataset that isn't LJSpeech would eventually cause a NaN loss at some point. The culprit appears to be the following line, which causes a division by zero if wav happens to contain perfect silence:

https://github.com/bshall/hifigan/blob/374a4569eae5437e2c80d27790ff6fede9fc1c46/hifigan/dataset.py#L106

I'm not sure what the best solution for this would be, as a quick fix I simply clipped the divisor so it can't reach zero:

wav = flip * gain * wav / max([wav.abs().max(), 0.001])
opened by cjay42 0
How to use this Vocoder with your Tacotron?

Thank you for your work. I used your Tacotron in your Universal Vocoding.The quality of the speech is excellent. However, the inference speed is slow. for that reason, I would like to use this hifigan as a vocoder. But Tacotron's n_mel is 80, while hifigan's n_mel is 128. How to use hifigan with Tacotron?

opened by gheyret 0

Releases(v0.1)

v0.1(Oct 17, 2021)

HiFi-GAN vocoders fine-tuned on the HuBERT-Soft and HuBERT-Discrete voice conversion systems.
Source code(tar.gz)
Source code(zip)
dev.txt(1.17 KB)
hifigan-67926ec6.pt(54.89 MB)
hifigan-hubert-discrete-bbad3043.pt(54.89 MB)
hifigan-hubert-soft-65f03469.pt(54.89 MB)
test.txt(1.17 KB)
train.txt(151.17 KB)

Owner

Benjamin van Niekerk

PhD student at Stellenbosch University. Interested in speech and audio technology.

GitHub Repository https://bshall.github.io/soft-vc/

Code for the paper "VisualBERT: A Simple and Performant Baseline for Vision and Language"

This repository contains code for the following two papers: VisualBERT: A Simple and Performant Baseline for Vision and Language (arxiv) with a short

463 Dec 09, 2022

Chainer Implementation of Semantic Segmentation using Adversarial Networks

Semantic Segmentation using Adversarial Networks Requirements Chainer (1.23.0) Differences Use of FCN-VGG16 instead of Dilated8 as Segmentor. Caution

99 Jun 28, 2022

an Evolutionary Algorithm assisted GAN

EvoGAN an Evolutionary Algorithm assisted GAN ckpts

3 Oct 09, 2022

Code for "PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clouds", CVPR 2021

PV-RAFT This repository contains the PyTorch implementation for paper "PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clou

43 Dec 05, 2022

Adversarial-Information-Bottleneck - Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneck (NeurIPS21)

NeurIPS 2021 Title: Distilling Robust and Non-Robust Features in Adversarial Exa

35 Dec 26, 2022

An 16kHz implementation of HiFi-GAN for soft-vc.

Related tags

Overview

HiFi-GAN

Example Usage

Train

Generate

Acknowledgements

You might also like...

Fast Soft Color Segmentation

Permute Me Softly: Learning Soft Permutations for Graph Representations

Multi-task Multi-agent Soft Actor Critic for SMAC

[ICLR 2022] Contact Points Discovery for Soft-Body Manipulations with Differentiable Physics

Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two

Implementation of TransGanFormer, an all-attention GAN that combines the finding from the recent GanFormer and TransGan paper

PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement.

This is a pytorch implementation of the NeurIPS paper GAN Memory with No Forgetting.

[CVPR 2021] Pytorch implementation of Hijack-GAN: Unintended-Use of Pretrained, Black-Box GANs

Comments

is pretrained weight of discriminator of base model available?

NaN during training when using own dataset

How to use this Vocoder with your Tacotron?

Releases(v0.1)

v0.1(Oct 17, 2021)

Owner

Benjamin van Niekerk

Code for the paper "VisualBERT: A Simple and Performant Baseline for Vision and Language"

Chainer Implementation of Semantic Segmentation using Adversarial Networks

an Evolutionary Algorithm assisted GAN

Code for "PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clouds", CVPR 2021

Adversarial-Information-Bottleneck - Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneck (NeurIPS21)

Implementation of Change-Based Exploration Transfer (C-BET)

Code for DeepXML: A Deep Extreme Multi-Label Learning Framework Applied to Short Text Documents

A fast, dataset-agnostic, deep visual search engine for digital art history

Exploring Simple 3D Multi-Object Tracking for Autonomous Driving (ICCV 2021)

Official PyTorch implementation of the paper "TEMOS: Generating diverse human motions from textual descriptions"

Deep learning for spiking neural networks

keyframes-CNN-RNN(action recognition)

chen2020iros: Learning an Overlap-based Observation Model for 3D LiDAR Localization.

Codes for paper "Towards Diverse Paragraph Captioning for Untrimmed Videos". CVPR 2021

An implementation of IMLE-Net: An Interpretable Multi-level Multi-channel Model for ECG Classification

Towards Flexible Blind JPEG Artifacts Removal (FBCNN, ICCV 2021)

DCSL - Generalizable Crowd Counting via Diverse Context Style Learning

DaReCzech is a dataset for text relevance ranking in Czech

Unet network with mean teacher for altrasound image segmentation

Freecodecamp Scientific Computing with Python Certification; Solution for Challenge 2: Time Calculator