Skip to content

Latest commit

 

History

History
126 lines (111 loc) · 7.09 KB

README.md

File metadata and controls

126 lines (111 loc) · 7.09 KB
Raw

FastSpeech2 + AISHELL-3 Voice Cloning (ECAPA-TDNN)

This example contains code used to train a FastSpeech2 model with AISHELL-3. The trained model can be used in Voice Cloning Task, We refer to the model structure of Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis. The general steps are as follows:

  1. Speaker Encoder: We use Speaker Verification to train a speaker encoder. Datasets used in this task are different from those used in FastSpeech2 because the transcriptions are not needed, we use more datasets, refer to ECAPA-TDNN.
  2. Synthesizer: We use the trained speaker encoder to generate speaker embedding for each sentence in AISHELL-3. This embedding is an extra input of FastSpeech2 which will be concated with encoder outputs.
  3. Vocoder: We use Parallel Wave GAN as the neural Vocoder, refer to voc1.

Dataset

Download and Extract

Download AISHELL-3 from it's Official Website and extract it to ~/datasets. Then the dataset is in the directory ~/datasets/data_aishell3.

Get MFA Result and Extract

We use MFA2.x to get durations for aishell3_fastspeech2. You can download from here aishell3_alignment_tone.tar.gz, or train your MFA model reference to mfa example (use MFA1.x now) of our repo.

Get Started

Assume the path to the dataset is ~/datasets/data_aishell3. Assume the path to the MFA result of AISHELL-3 is ./aishell3_alignment_tone.

Run the command below to

  1. source path.
  2. preprocess the dataset.
  3. train the model.
  4. synthesize waveform from metadata.jsonl.
  5. start a voice cloning inference.
./run.sh

You can choose a range of stages you want to run, or set stage equal to stop-stage to use only one stage, for example, running the following command will only preprocess the dataset.

./run.sh --stage 0 --stop-stage 0

Data Preprocessing

CUDA_VISIBLE_DEVICES=${gpus} ./local/preprocess.sh ${conf_path}

When it is done. A dump folder is created in the current directory. The structure of the dump folder is listed below.

dump
├── dev
│   ├── norm
│   └── raw
├── embed
│   ├── SSB0005
│   ├── SSB0009
│   ├── ...
│   └── ...
├── phone_id_map.txt
├── speaker_id_map.txt
├── test
│   ├── norm
│   └──  raw
└── train
    ├── energy_stats.npy
    ├── norm
    ├── pitch_stats.npy
    ├── raw
    └── speech_stats.npy

The embed contains the generated speaker embedding for each sentence in AISHELL-3, which has the same file structure with wav files and the format is .npy.

The computing time of utterance embedding can be x hours.

The dataset is split into 3 parts, namely train, dev, and test, each of which contains a norm and raw subfolder. The raw folder contains speech、pitch and energy features of each utterance, while the norm folder contains normalized ones. The statistics used to normalize features are computed from the training set, which is located in dump/train/*_stats.npy.

Also, there is a metadata.jsonl in each subfolder. It is a table-like file that contains phones, text_lengths, speech_lengths, durations, the path of speech features, the path of pitch features, the path of energy features, speaker, and id of each utterance.

The preprocessing step is very similar to that one of tts3, but there is one more ECAPA-TDNN/inference step here.

Model Training

./local/train.sh calls ${BIN_DIR}/train.py.

CUDA_VISIBLE_DEVICES=${gpus} ./local/train.sh ${conf_path} ${train_output_path}

The training step is very similar to that one of tts3, but we should set --voice-cloning=True when calling ${BIN_DIR}/train.py.

Synthesizing

We use parallel wavegan as the neural vocoder. Download pretrained parallel wavegan model from pwg_aishell3_ckpt_0.5.zip and unzip it.

unzip pwg_aishell3_ckpt_0.5.zip

Parallel WaveGAN checkpoint contains files listed below.

pwg_aishell3_ckpt_0.5
├── default.yaml                   # default config used to train parallel wavegan
├── feats_stats.npy                # statistics used to normalize spectrogram when training parallel wavegan
└── snapshot_iter_1000000.pdz      # generator parameters of parallel wavegan

./local/synthesize.sh calls ${BIN_DIR}/../synthesize.py, which can synthesize waveform from metadata.jsonl.

CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}

The synthesizing step is very similar to that one of tts3, but we should set --voice-cloning=True when calling ${BIN_DIR}/../synthesize.py.

Voice Cloning

Assume there are some reference audios in ./ref_audio (the format must be wav here)

ref_audio
├── 001238.wav
├── LJ015-0254.wav
└── audio_self_test.wav

./local/voice_cloning.sh calls ${BIN_DIR}/../voice_cloning.py

CUDA_VISIBLE_DEVICES=${gpus} ./local/voice_cloning.sh ${conf_path} ${train_output_path} ${ckpt_name} ${ref_audio_dir}

Pretrained Model

Model Step eval/loss eval/l1_loss eval/duration_loss eval/pitch_loss eval/energy_loss
default 2(gpu) x 96400 0.991855 0.599517 0.052142 0.094877 0.245318

FastSpeech2 checkpoint contains files listed below. (There is no need for speaker_id_map.txt here )

fastspeech2_aishell3_ckpt_vc2_1.2.0
├── default.yaml            # default config used to train fastspeech2
├── energy_stats.npy        # statistics used to normalize energy when training fastspeech2
├── phone_id_map.txt        # phone vocabulary file when training fastspeech2
├── pitch_stats.npy         # statistics used to normalize pitch when training fastspeech2
├── snapshot_iter_96400.pdz # model parameters and optimizer states
└── speech_stats.npy        # statistics used to normalize spectrogram when training fastspeech2