"""
This recipe supports two corpora: LibriTTS and LibriTTS-R.
---
LibriTTS is a multi-speaker English corpus of approximately 585 hours of read English speech at 24kHz sampling rate, prepared by Heiga Zen with the assistance of Google Speech and Google Brain team members. The LibriTTS corpus is designed for TTS research. It is derived from the original materials (mp3 audio files from LibriVox and text files from Project Gutenberg) of the LibriSpeech corpus. The main differences from the LibriSpeech corpus are listed below:
The audio files are at 24kHz sampling rate.
The speech is split at sentence breaks.
Both original and normalized texts are included.
Contextual information (e.g., neighbouring sentences) can be extracted.
Utterances with significant background noise are excluded.
For more information, refer to the paper "LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech", Heiga Zen, Viet Dang, Rob Clark, Yu Zhang, Ron J. Weiss, Ye Jia, Zhifeng Chen, and Yonghui Wu, arXiv, 2019. If you use the LibriTTS corpus in your work, please cite this paper where it was introduced.
---
LibriTTS-R [1] is a sound quality improved version of the LibriTTS corpus (http://www.openslr.org/60/) which is a multi-speaker English corpus of approximately 585 hours of read English speech at 24kHz sampling rate, published in 2019. The constituent samples of LibriTTS-R are identical to those of LibriTTS, with only the sound quality improved. To improve sound quality, a speech restoration model, Miipher proposed by Yuma Koizumi [2], was used.
For more information, refer to the paper [1]. If you use the LibriTTS-R corpus in your work, please cite the dataset paper [1] where it was introduced.
Audio samples of the ground-truth and TTS generated samples are available at the demo page: https://google.github.io/df-conformer/librittsr/
[1] Yuma Koizumi, Heiga Zen, Shigeki Karita, Yifan Ding, Kohei Yatabe, Nobuyuki Morioka, Michiel Bacchiani, Yu Zhang, Wei Han, and Ankur Bapna, "LibriTTS-R: A Restored Multi-Speaker Text-to-Speech Corpus," arXiv, 2023.
[2] Yuma Koizumi, Heiga Zen, Shigeki Karita, Yifan Ding, Kohei Yatabe, Nobuyuki Morioka, Yu Zhang, Wei Han, Ankur Bapna, and Michiel Bacchiani, "Miipher: A Robust Speech Restoration Model Integrating Self-Supervised Speech and Text Representations," arXiv, 2023.
"""
import logging
import re
import shutil
import tarfile
from pathlib import Path
from typing import Dict, Optional, Sequence, Union
from tqdm import tqdm
from lhotse import (
RecordingSet,
SupervisionSegment,
SupervisionSet,
fix_manifests,
validate_recordings_and_supervisions,
)
from lhotse.recipes.utils import manifests_exist, read_manifests_if_cached
from lhotse.utils import Pathlike, resumable_download, safe_extract
LIBRITTS = (
"dev-clean",
"dev-other",
"test-clean",
"test-other",
"train-clean-100",
"train-clean-360",
"train-other-500",
)
[docs]
def download_librittsr(
target_dir: Pathlike = ".",
dataset_parts: Optional[Union[str, Sequence[str]]] = "all",
force_download: Optional[bool] = False,
base_url: Optional[str] = "http://www.openslr.org/resources",
) -> Path:
"""
Download and untar the LibriTTS-R dataset.
:param target_dir: Pathlike, the path of the dir to storage the dataset.
:param dataset_parts: "all", or a list of splits (e.g. "dev-clean") to download.
:param force_download: Bool, if True, download the tars no matter if the tars exist.
:param base_url: str, the url of the OpenSLR resources.
:return: the path to downloaded and extracted directory with data.
"""
return download_libritts(
target_dir=target_dir,
dataset_parts=dataset_parts,
force_download=force_download,
base_url=base_url,
use_librittsr=True,
)
[docs]
def download_libritts(
target_dir: Pathlike = ".",
use_librittsr: bool = False,
dataset_parts: Optional[Union[str, Sequence[str]]] = "all",
force_download: Optional[bool] = False,
base_url: Optional[str] = "http://www.openslr.org/resources",
) -> Path:
"""
Download and untar the LibriTTS dataset.
:param target_dir: Pathlike, the path of the dir to storage the dataset.
:param use_librittsr: Bool, if True, we'll download the LibriTTS-R dataset instead.
:param dataset_parts: "all", or a list of splits (e.g. "dev-clean") to download.
:param force_download: Bool, if True, download the tars no matter if the tars exist.
:param base_url: str, the url of the OpenSLR resources.
:return: the path to downloaded and extracted directory with data.
"""
target_dir = Path(target_dir)
target_dir.mkdir(parents=True, exist_ok=True)
if dataset_parts == "all" or dataset_parts[0] == "all":
dataset_parts = LIBRITTS
elif isinstance(dataset_parts, str):
dataset_parts = [dataset_parts]
if use_librittsr:
name = "LibriTTS_R"
openslr_corpus_id = "141"
else:
name = "LibriTTS"
openslr_corpus_id = "60"
for part in tqdm(dataset_parts, desc=f"Downloading {name} parts"):
if part not in LIBRITTS:
logging.warning(
f"Skipping invalid dataset part name: {part} (possible choices: {LIBRITTS})"
)
url = f"{base_url}/{openslr_corpus_id}"
tar_name = f"{part}.tar.gz"
tar_path = target_dir / tar_name
part_dir = target_dir / f"{name}/{part}"
completed_detector = part_dir / ".completed"
if completed_detector.is_file():
logging.info(f"Skipping {part} because {completed_detector} exists.")
continue
full_url = f"{url}/{tar_name}"
if use_librittsr:
full_url = full_url.replace("-", "_")
resumable_download(full_url, filename=tar_path, force_download=force_download)
shutil.rmtree(part_dir, ignore_errors=True)
with tarfile.open(tar_path) as tar:
safe_extract(tar, path=target_dir)
completed_detector.touch()
return target_dir
[docs]
def prepare_libritts(
corpus_dir: Pathlike,
dataset_parts: Union[str, Sequence[str]] = "all",
output_dir: Optional[Pathlike] = None,
num_jobs: int = 1,
link_previous_utt: bool = False,
) -> Dict[str, Dict[str, Union[RecordingSet, SupervisionSet]]]:
"""
Returns the manifests which consist of the Recordings and Supervisions.
When all the manifests are available in the ``output_dir``, it will simply read and return them.
:param corpus_dir: Pathlike, the path of the data dir.
:param dataset_parts: string or sequence of strings representing dataset part names, e.g. 'train-clean-100', 'train-clean-5', 'dev-clean'.
By default we will infer which parts are available in ``corpus_dir``.
:param output_dir: Pathlike, the path where to write the manifests.
:param num_jobs: the number of parallel workers parsing the data.
:param link_previous_utt: If true adds previous utterance id to supervisions.
Useful for reconstructing chains of utterances as they were read.
If previous utterance was skipped from LibriTTS datasets previous_utt label is None.
:return: a Dict whose key is the dataset part, and the value is Dicts with the keys 'audio' and 'supervisions'.
"""
corpus_dir = Path(corpus_dir)
assert corpus_dir.is_dir(), f"No such directory: {corpus_dir}"
if dataset_parts == "all" or dataset_parts[0] == "all":
dataset_parts = LIBRITTS
elif isinstance(dataset_parts, str):
assert dataset_parts in LIBRITTS
dataset_parts = [dataset_parts]
manifests = {}
if output_dir is not None:
output_dir = Path(output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
# Maybe the manifests already exist: we can read them and save a bit of preparation time.
manifests = read_manifests_if_cached(
dataset_parts=dataset_parts, output_dir=output_dir, prefix="libritts"
)
# Contents of the file
# ;ID |SEX| SUBSET |MINUTES| NAME
# 14 | F | train-clean-360 | 25.03 | ...
# 16 | F | train-clean-360 | 25.11 | ...
# 17 | M | train-clean-360 | 25.04 | ...
spk2gender = {}
if (corpus_dir / "SPEAKERS.txt").is_file():
spk2gender = {
spk_id.strip(): gender.strip()
for spk_id, gender, *_ in (
line.split("|")
for line in (corpus_dir / "SPEAKERS.txt").read_text().splitlines()
if not line.startswith(";")
)
}
for part in tqdm(dataset_parts, desc="Preparing LibriTTS parts"):
if manifests_exist(part=part, output_dir=output_dir, prefix="libritts"):
logging.info(f"LibriTTS subset: {part} already prepared - skipping.")
continue
part_path = corpus_dir / part
# We are ignoring weird files such as ._84_121550_000007_000000.wav
# Maybe LibriTTS-R will fix it in later distributions.
# Also, the file 1092_134562_000013_000004.wav is corrupted as of May 31st.
recordings = RecordingSet.from_dir(
part_path,
"*.wav",
num_jobs=num_jobs,
exclude_pattern=r"^(\._.+|1092_134562_000013_000004\.wav)$",
)
supervisions = []
for trans_path in tqdm(
part_path.rglob("*.trans.tsv"),
desc="Scanning transcript files (progbar per speaker)",
leave=False,
):
if re.match(r"^\._.+$", trans_path.name) is not None:
continue
# The trans.tsv files contain only the recordings that were kept for LibriTTS.
# Example path to a file:
# /export/corpora5/LibriTTS/dev-clean/84/121123/84_121123.trans.tsv
#
# Example content:
# 84_121123_000007_000001 Maximilian. Maximilian.
# 84_121123_000008_000000 Villefort rose, half ashamed of being surprised in such a paroxysm of grief. Villefort rose, half ashamed of being surprised in such a paroxysm of grief.
# book.tsv contains additional metadata
utt2snr = [
(rec_id, float(snr))
for rec_id, *_, snr in map(
str.split,
(
trans_path.parent
/ trans_path.name.replace(".trans.tsv", ".book.tsv")
)
.read_text()
.splitlines(),
)
]
# keeps the order of uttids as they appear in book.tsv
uttids = [r for r, _ in utt2snr]
utt2snr = dict(utt2snr)
if link_previous_utt:
# Using the property of sorted keys to find previous utterance
# The keys has structure speaker_book_x_y e.g. 1089_134691_000004_000001
utt2prevutt = dict(zip(uttids + [None], [None] + uttids))
prev_rec_id = None
for line in trans_path.read_text().splitlines():
rec_id, orig_text, norm_text = line.split("\t")
if rec_id not in recordings:
logging.warning(
f"No recording exists for utterance id {rec_id}, skipping (in {trans_path})"
)
continue
spk_id = rec_id.split("_")[0]
customd = {"orig_text": orig_text, "snr": utt2snr.get(rec_id)}
if link_previous_utt:
# all recordings ids should be in the book.csv
# but they are some missing e.g. 446_123502_000030_000003
prev_utt = utt2prevutt.get(rec_id, None)
# previous utterance has to be present in trans.csv - otherwise it was skipped
prev_utt = prev_utt if prev_utt == prev_rec_id else None
customd["prev_utt"] = prev_utt
prev_rec_id = rec_id
supervisions.append(
SupervisionSegment(
id=rec_id,
recording_id=rec_id,
start=0.0,
duration=recordings[rec_id].duration,
channel=0,
text=norm_text,
language="English",
speaker=spk_id,
gender=spk2gender.get(spk_id),
custom=customd,
)
)
supervisions = SupervisionSet.from_segments(supervisions)
recordings, supervisions = fix_manifests(recordings, supervisions)
validate_recordings_and_supervisions(recordings, supervisions)
if output_dir is not None:
supervisions.to_file(output_dir / f"libritts_supervisions_{part}.jsonl.gz")
recordings.to_file(output_dir / f"libritts_recordings_{part}.jsonl.gz")
manifests[part] = {"recordings": recordings, "supervisions": supervisions}
return manifests
prepare_librittsr = prepare_libritts