Direct Download Links

New OPUS direct download links

OPUS torrent micro release

OPUS torrent micro release

• Dataset conversion to OPUS
• OPUS torrent - https://academictorrents.com/details/95b4cab0f99850e119114c8b6df00193ab5fa34f
• OPUS helpers and build instructions - https://github.com/snakers4/open_stt/#how-to-open-opus
• Coming soon - new unlimited direct links
• Further reading links

Finally a v1.0 release with 3x more data

The largest Russian STT dataset up-to-date

• ~16m utterances;
• ~20 000 hours;
• 2,3 TB of data(in .wav format in int16);
• A wide variety of practical, close to real-life domains;

Major highlights

• ~3 000 hours of a completely new domain - public speech;
• A huge Radio dataset update with 10 000+ hours ;
• A 5% demo version of new Radio/Public Speech datasets;
• Vastly improved dataset normalization;
• Overall annotation quality is improved:
  • Upstream model quality improvement;
  • No more "dangling" letters;
  • Improved voice activity detection;
    See the above TLDR bullets;

Next steps

• Major past error clean-up planned in 1.1;
• Refine and publish speaker labels, probably add speakers for old datasets;
• Improve / re-upload some of the existing datasets, refine the STT labels;
• Probably add new languages;
• Add pre-trained models;

New major release - radio / youtube / data quality distillation

TLDR:

• 855 GB (in .wav format in int16) non archived;
• (new!) A new domain - radio;
• (new!) A larger YouTube dataset with 1000+ additional hours;
• (new!) A small (300 hours) YouTube dataset downloaded in maximum quality;
• (new!) 18 hours in 3 validation sets for YouTube / books / public calls with ground truth annotations;
• See the distilled files with "bad" data in this issue;

Added full WAV torrent release

• An MP3 version of the dataset;
• A WAV version of the dataset;

Fixed issues with no txt files in torrents

Added txt files to torrents and direct archives.
Updated torrents.

Added torrent

Added link to a torrent download.

Dataset conversion to MP3

Key changes:

• Converted the majority of the dataset to MP3;
• Added download script, md5 hashes into download script;
• Fixed license;
• Added items to FAQ and common issues;

THE MAJORITY OF WAV LINKS WILL BE DELETED SOON.

Coming soon:

• Download via torrent;
• Large (1,500 hours) YouTube dataset;
• ... and more)

Dataset composition

Dataset	Utterances	Hours	GB	Av s/chars	Comment	Annotation	Quality/noise
public_youtube1500 (*)		1,500			* Coming soon
audiobook_2	1,149,404	1,511	166	4.7s / 56	Books	Alignment (*)	95% / crisp
public_youtube700	759,483	701	75	3.3s / 43	Youtube videos	Subtitles	95% / ~crisp
tts_russian_addresses	1,741,838	754	81	1.6s / 20	Russian addresses	TTS 4 voices	100% / crisp
asr_public_phone_calls_2	603,797	601	66	3.6s / 37	Phone calls	ASR	70% / noisy
asr_public_phone_calls_1	233,868	211	23	3.3s / 29	Phone calls	ASR	70% / noisy
asr_public_stories_2	78,186	78	9	3.5s / 43	Books	ASR	80% / crisp
asr_public_stories_1	46,142	38	4	3.0s / 30	Books	ASR	80% / crisp
public_series_1	20,243	17	2	3.1s / 38	Youtube videos	Subtitles	95% / ~crisp
ru_RU	5,826	17	2	11s / 12	Public dataset	Alignment	99% / crisp
voxforge_ru	8,344	17	2	7.5s / 77	Public dataset	Reading	100% / crisp
russian_single	3,357	9	1	9.3s / 102	Public dataset	Alignment	99% / crisp
public_lecture_1	6,803	6	1	3.4s / 47	Lectures	Subtitles	95% / crisp
Total	4,657,291	3,961	431

Links

Meta data file.

Dataset	GB, wav	GB, mp3	Wav	Mp3	Source	Manifest
audiobook_2	166	21.0	down	part1	Sources from the Internet + alignment	link
asr_public_phone_calls_2	66	7.5	down	part1	Sources from the Internet + ASR	link
asr_public_stories_2	9 (7.5)	NA	part1	NA	Sources from the Internet + alignment	link
tts_russian_addresses_rhvoice_4voices	80.9	9.9	down	part1	TTS	link
public_youtube700	75.0	9.6	down	part1	YouTube videos	link
asr_public_phone_calls_1	22.7	2.6	down	part1	Sources from the Internet + ASR	link
asr_public_stories_1	4.1	0.5	down	part1	Public stories	link
public_series_1	1.9	0.2	down	part1	Public series	link
ru_RU	1.9	0.2	down	part1	Caito.de dataset	link
voxforge_ru	1.9	0.2	down	part1	Voxforge dataset	link
russian_single	0.9	0.1	down	part1	Russian single speaker dataset	link
public_lecture_1	0.7	0.1	down	part1	Sources from the Internet	link
Total	431	52

Added large audio book corpus, large phone call database, asr stories

Key changes:

• Added dataset: 1500 hours of aligned books, 600+ hours of phone calls, 78 hours of ASR stories.
• Formatting changes;
• Added license;
• Added items to FAQ and common issues;

Coming soon:

• Large (1,500 hours) YouTube dataset;
• ... and more)

Dataset composition

Dataset	Utterances	Hours	GB	Av s/chars	Comment	Annotation	Quality/noise
public_youtube1500 (*)		1,500			* Coming soon
audiobook_2	1,149,404	1,511	166	4.7s / 56	Books	Alignment	99% / crisp
audiobook_1	196,666	237	26	4.3s / 50	Books	Alignment	99% / crisp
public_youtube700	759,483	701	75	3.3s / 43	Youtube videos	Subtitles	95% / ~crisp
tts_russian_addresses	1,741,838	754	81	1.6s / 20	Russian addresses	TTS 4 voices	100% / crisp
asr_public_phone_calls_2	603,797	601	66	3.6s / 37	Phone calls	ASR	70% / noisy
asr_public_phone_calls_1	233,868	211	23	3.3s / 29	Phone calls	ASR	70% / noisy
asr_public_stories_2	78,186	78	9	3.5s / 43	Books	ASR	80% / crisp
asr_public_stories_1	46,142	38	4	3.0s / 30	Books	ASR	80% / crisp
public_series_1	20,243	17	2	3.1s / 38	Youtube videos	Subtitles	95% / ~crisp
ru_RU	5,826	17	2	11s / 12	Public dataset	Alignment	99% / crisp
voxforge_ru	8,344	17	2	7.5s / 77	Public dataset	Reading	100% / crisp
russian_single	3,357	9	1	9.3s / 102	Public dataset	Alignment	99% / crisp
public_lecture_1	6,803	6	1	3.4s / 47	Lectures	Subtitles	95% / crisp
Total	4,853,957	4,198	457

Links

Meta data file.

Dataset	GB	GB, compressed	Audio	Source	Manifest
audiobook_1	26	20.8	part1	Public books + alignment	link
audiobook_2	166	131.7	part1, part2, part3, part4, part5, part6, part7	Public books + alignment	link
asr_public_phone_calls_2	66	51.7	part1, part2, part3	ASR + public phone calls	link
asr_public_stories_2	9	7.5	part1	Public books + alignment	link
tts_russian_addresses_rhvoice_4voices	80.9	67.0	part1, part2, part3, part4	TTS	link
public_youtube700	75.0	67.0	part1, part2, part3, part4	YouTube videos	link
asr_public_phone_calls_1	22.7	19.0	part1	ASR + public phone calls	link
asr_public_stories_1	4.1	3.8	part1	Public stories	link
public_series_1	1.9	1.7	part1	Public series	link
ru_RU	1.9	1.4	part1	Caito.de dataset	link
voxforge_ru	1.9	1.5	part1	Voxforge dataset	link
russian_single	0.9	0.7	part1	Russian single speaker dataset	link
public_lecture_1	0.7	0.6	part1	Public lectures	link
Total	190	163

Check md5sum

md5sum /path/to/downloaded/file

Click to expand

type	md5sum	file
manifest	b0ce7564ba90b121aeb13aada73a6e30	asr_public_phone_calls_1.csv
manifest	6867d14dfdec1f9e9b8ca2f1de9ceda6	asr_public_phone_calls_2.csv
manifest	0bdd77e15172e654d9a1999a86e92c7f	asr_public_stories_1.csv
manifest	f388013039d94dc36970547944db51c7	asr_public_stories_2.csv
manifest	697738331b6021890c29a0d415d0f22d	private_buriy_audiobooks_1.csv
manifest	3b67e27c1429593cccbf7c516c4b582d	private_buriy_audiobooks_2.csv
manifest	04027c20eb3aff05f6067957ecff856b	public_lecture_1.csv
manifest	89da3f1b6afcd4d4936662ceabf3033e	public_series_1.csv
manifest	a81dfb018c88d0ecd5194ab3d8ff6c95	public_youtube700.csv
manifest	c858f020729c34ba0ab525bbb8950d0c	ru_RU.csv
manifest	0275525914825dec663fd53390fdc9a0	russian_single.csv
manifest	52f406f4e30fcc8c634f992befd91beb	tts_russian_addresses_rhvoice_4voices.csv
audio	a5496898ee78654bf398ec6df71540d7	asr_public_phone_calls_1.tar.gz
audio	e4df5ef50787384648b59f5a87edc0c6	asr_public_phone_calls_2.tar.gz
audio	97594127a922df8a7bcc2eecd2470805	asr_public_phone_calls_2.tar.gz_aa
audio	f9b6475f0f2898b16d9e6e0e648fb531	asr_public_...

Added medium-sized YouTube dataset and TTS dataset

Added medium-sized YouTube dataset and TTS dataset

Key changes:

• The storage format was changed to on-disk DB with hashes;
• Added a 700 hour YouTube dataset;
• Added a 700+ hour TTS dataset with Russian addresses;
• Added some utils to work with manifests;
• Added manifest files for easier porting into your ASR application;
• Discarded previous links;
• Dataset format will be uniform from now, new "datasets" will be just added;

Coming soon:

• Large (1,500 hours) phone call dataset;
• Large (1,500 hours) YouTube dataset;
• ... and more)

Dataset composition

Dataset	Utterances	Hours	GB	Av len/chars	Comment	Annotation	Quality/noise
asr_public_phone_calls_2 (*)		1,500			* Coming soon
public_youtube1500 (*)		1,500			* Coming soon
tts_russian_addresses	1,741,838	754	81	1.6s / 20	Russian addresses	TTS, 4 voices	100% / crisp
public_youtube700	759,483	701	75	3.3s / 43	Youtube videos	Subtitles	>95% / ~crisp
asr_public_phone_calls_1	233,868	211	23	3.3s / 29	Phone calls	ASR	70% / noisy
asr_public_stories_1	46,142	38	4	3.0s / 30	Books	ASR	70% / crisp
public_series_1	20,243	17	2	3.1s / 38	Youtube videos	Subtitles	95% / ~crisp
ru_RU	5,826	17	2	10.8s / 12	Public dataset	Alignment	99% / crisp
voxforge_ru	8,344	17	2	7.5s / 77	Public dataset	Reading	100% / crisp
russian_single	3,357	9	1	9.3s / 102	Public dataset	Alignment	99% / crisp
public_lecture_1	6,803	6	1	3.4s / 47	Lectures	Subtitles	>95% / crisp
Total	2,825,904	1,771	190

Links

Meta data file.

Dataset	GB	GB, compressed	Audio	Source	Manifest
tts_russian_addresses_rhvoice_4voices	80.9	67.0	part1, part2, part3, part4	TTS	link
public_youtube700	75.0	67.0	part1, part2, part3, part4	YouTube videos	link
asr_public_phone_calls_1	22.7	19.0	part1	ASR + public phone calls	link
asr_public_stories_1	4.1	3.8	part1	Public stories	link
public_series_1	1.9	1.7	part1	Public series	link
ru_RU	1.9	1.4	part1	Caito.de dataset	link
voxforge_ru	1.9	1.5	part1	Voxforge dataset	link
russian_single	0.9	0.7	part1	Russian single speaker dataset	link
public_lecture_1	0.7	0.6	part1	Public lectures	link
Total	190	163