Vocoder with HiFIGAN Unit

Work In Progress ....

Install SpeechBrain

First of all, please install tranformers and SpeechBrain with the following command:

pip install speechbrain transformers

Please notice that we encourage you to read our tutorials and learn more about SpeechBrain.

Using the Vocoder

import torchaudio
from speechbrain.inference.encoders import MelSpectrogramEncoder
from speechbrain.inference.vocoders import UnitHIFIGAN
from speechbrain.lobes.models.huggingface_transformers.discrete_hubert import (
    DiscreteHuBERT,
)

speaker_encoder_source = "speechbrain/spkrec-ecapa-voxceleb-mel-spec"
speech_encoder_source = "facebook/hubert-base-ls960"
kmeans_folder = "speechbrain/SSL_Quantization"
kmeans_filename = "LJSpeech_hubert_k128_L7.pt"
layer = 7
vocoder_source = "chaanks/hifigan-unit-hubert-l7-k128-ljspeech-libritts"
save_path = "tmpdir"
device = "cuda"
sample_rate = 16000

wav = "chaanks/hifigan-unit-hubert-l7-k128-ljspeech-libritts/test.wav"

speaker_encoder = MelSpectrogramEncoder.from_hparams(
    source=speaker_encoder_source,
    run_opts={"device": str(device)},
    savedir=save_path + "/spk_encoder",
)

speech_encoder = DiscreteHuBERT(
    source=speech_encoder_source,
    save_path=save_path + "/speech_encoder",
    kmeans_filename=kmeans_filename,
    kmeans_cache_dir=save_path + "/kmeans",
    kmeans_repo_id=kmeans_folder,
    output_norm=False,
    freeze=True,
    freeze_feature_extractor=True,
    apply_spec_augment=False,
    output_all_hiddens=True,
    ssl_layer_num=layer,
).to(device)

vocoder = UnitHIFIGAN.from_hparams(
    source=vocoder_source,
    run_opts={"device": str(device)},
    savedir=save_path + "/vocoder",
)

audio = speaker_encoder.load_audio(wav)
audio = audio.to(device)
spk = speaker_encoder.encode_waveform(audio)
_, codes = speech_encoder(audio.unsqueeze(0))
waveform = vocoder.decode_unit(codes.squeeze(0), spk=spk.reshape(-1))
torchaudio.save("test.wav", waveform.cpu(), sample_rate=sample_rate)

Inference on GPU

To perform inference on the GPU, add run_opts={"device":"cuda"} when calling the from_hparams method.

Limitations

The SpeechBrain team does not provide any warranty on the performance achieved by this model when used on other datasets.

Referencing SpeechBrain

@misc{SB2021,
    author = {Ravanelli, Mirco and Parcollet, Titouan and Rouhe, Aku and Plantinga, Peter and Rastorgueva, Elena and Lugosch, Loren and Dawalatabad, Nauman and Ju-Chieh, Chou and Heba, Abdel and Grondin, Francois and Aris, William and Liao, Chien-Feng and Cornell, Samuele and Yeh, Sung-Lin and Na, Hwidong and Gao, Yan and Fu, Szu-Wei and Subakan, Cem and De Mori, Renato and Bengio, Yoshua },
    title = {SpeechBrain},
    year = {2021},
    publisher = {GitHub},
    journal = {GitHub repository},
    howpublished = {\\\\url{https://github.com/speechbrain/speechbrain}},
  }

About SpeechBrain

SpeechBrain is an open-source and all-in-one speech toolkit. It is designed to be simple, extremely flexible, and user-friendly. Competitive or state-of-the-art performance is obtained in various domains.

Website: https://speechbrain.github.io/

GitHub: https://github.com/speechbrain/speechbrain