Loading datasets...
Task: ASR
Release Date: 6/17/2026
Format: MP3
Size: 5.90 GB
Share
A collection of read speech recordings in Latvian (Latviešu).
Restrictions/Special Constraints
None provided.
Forbidden Usage
It is forbidden to attempt to determine the identity of speakers in the Common Voice datasets. It is forbidden to re-host or re-share this dataset.
Intended Use
This dataset is intended to be used for training and evaluating automatic speech recognition (ASR) models. It may also be used for applications relating to computer-aided language learning (CALL) and language or heritage revitalisation.
lv)This datasheet is for cv-corpus-26.0-2026-06-12 of the Mozilla Common Voice Scripted Speech dataset for Latvian [Latviešu - lv]. The dataset contains 233550 clips representing 312.63 hours of recorded speech (266.89 hours validated) from 6557 speakers, recorded from a text corpus of 99,616 sentences.
Latvian is an Baltic language belonging to the Indo-European language family. It is spoken in the Baltic region, and is the official language of Latvia as well as one of the official languages of the European Union. There are about 1.5 million native Latvian speakers in Latvia and 100,000 abroad. It uses the Latin alphabet with additional diacritical marks to represent specific sounds.
There are slight stylistic speach variations among different regions of Latvia, but they do not affect mutual
intelligibility. Latgale region has more significant differences with main Latvian language. It is sometimes
considered a dialect, but Latgalian (ltg) speech is collected as a separate language in Common Voice.
| Code | Accent | Clips | Speakers |
|---|---|---|---|
| - | 55,761 (23.9%) | 1,184 (18.1%) |
The dataset includes the following self-declared age and gender distributions. A coverage summary is shown below each table.
Self-declared gender information. The table shows clip and speaker counts with percentages. Speakers who did not declare a gender are listed as Unspecified. A dash (-) indicates zero.
| Code | Gender | Clips | Speakers |
|---|---|---|---|
| male_masculine | Male, masculine | 57,568 (24.6%) | 366 (5.6%) |
| female_feminine | Female, feminine | 71,132 (30.5%) | 514 (7.8%) |
| transgender | Transgender | - | - |
| non-binary | Non-binary | 114 (0.0%) | 1 (0.0%) |
| do_not_wish_to_say | Prefer not to say | 249 (0.1%) | 2 (0.0%) |
| - | Unspecified | 104,487 (44.7%) | 6,226 (95.0%) |
Gender declared: 129,063 of 233,550 clips (55.3%), 331 of 6,557 speakers (5.0%)
Self-declared age information. The table shows clip and speaker counts with percentages. Speakers who did not declare an age are listed as Unspecified. A dash (-) indicates zero.
| Code | Age | Clips | Speakers |
|---|---|---|---|
| teens | Teens | 3,081 (1.3%) | 84 (1.3%) |
| twenties | Twenties | 15,952 (6.8%) | 173 (2.6%) |
| thirties | Thirties | 66,057 (28.3%) | 294 (4.5%) |
| fourties | Fourties | 21,662 (9.3%) | 228 (3.5%) |
| fifties | Fifties | 19,177 (8.2%) | 101 (1.5%) |
| sixties | Sixties | 4,396 (1.9%) | 33 (0.5%) |
| seventies | Seventies | 220 (0.1%) | 6 (0.1%) |
| eighties | Eighties | 135 (0.1%) | 2 (0.0%) |
| nineties | Nineties | - | - |
| - | Unspecified | 102,870 (44.0%) | 6,207 (94.7%) |
Age declared: 130,680 of 233,550 clips (56.0%), 350 of 6,557 speakers (5.3%)
Clip buckets
| Bucket | Clips |
|---|---|
| Validated | 199,378 (85.4%) |
| Invalidated | 11,783 (5.0%) |
| Other | 22,389 (9.6%) |
Training splits
| Split | Clips |
|---|---|
| Train | 14,951 (7.5%) |
| Dev | 7,875 (3.9%) |
| Test | 7,875 (3.9%) |
Training split coverage: 30,701 of 199,378 validated clips (15.4%)
The dataset contains 199378 validated, 11783 invalidated, and 22389 unresolved clips. The average clip duration is 4.819 seconds.
Validated sentences: 34,642
| Category | Count |
|---|---|
| Unvalidated sentences | 64,974 |
| Pending sentences | 62,985 |
| Rejected sentences | 1,989 |
| Reported sentences | 3,139 |
The corpus contains 99,616 sentences: 34,642 validated and 64,974 unvalidated (62,985 pending review, 1,989 rejected), with 3,139 reported for review.
Latvian alphabet is based on the Latin script and consists of 33 letters, including special characters with diacritical marks such as ā, č, ē, ģ, ī, ķ, ļ, ņ, š, ū, and ž.
Historic texts also used letter /ŗ/, but use of it has been officially discontinued since 1946. This letter is equivalent to letter /r/ in modern Latvian orthography.
Letter /o/ can represent three different sounds [ua̯], [ɔ], [ɔː].
Letter /e/ can represent two different sounds [e], [æ].
a ā b c č d e ē f g ģ h i ī j k ķ l ļ m n ņ o p r s š t u ū v z ž
There follows a randomly selected sample of five sentences from the corpus.
Piesedz to stūri!
Un vispār, kas tev ļāva lemt par mums?
Guļu gultā, ierakstu un klausos tekstus.
Labi. Dod man dažas stundas.
Tāpēc, ja neesi apmierināts ar manu progresu, kāpēc, pie velna, nevācies ārā?
Text sources in the dataset consists of user added content as well as balanced sets of samples from Wikipedia, European parliament transcripts, out of copyright books and other sources.
| Source | Sentences |
|---|---|
| https://korpuss.lv/id/PanDi | 10,134 (29.3%) |
| sentence-collector | 8,275 (23.9%) |
| https://lv.wikipedia.org/ | 3,631 (10.5%) |
| https://korpuss.lv/id/Saeima | 2,521 (7.3%) |
| https://huggingface.co/datasets/open_subtitles | 2,337 (6.7%) |
| Autobiogrāfijas | 2,322 (6.7%) |
| Pašatsauce | 1,399 (4.0%) |
| https://huggingface.co/datasets/europarl_bilingual | 1,224 (3.5%) |
| https://huggingface.co/datasets/clips/mqa/viewer/lv-all-question/train | 619 (1.8%) |
| Other | 2,180 (6.3%) |
| Code | Domain | Clips | Speakers |
|---|---|---|---|
| general | General | 787 (0.3%) | 288 (4.4%) |
| agriculture_food | Agriculture and Food | 9 (0.0%) | 9 (0.1%) |
| automotive_transport | Automotive and Transport | 3 (0.0%) | 3 (0.0%) |
| finance | Finance | 5 (0.0%) | 5 (0.1%) |
| service_retail | Service and Retail | 17 (0.0%) | 16 (0.2%) |
| healthcare | Healthcare | 36 (0.0%) | 31 (0.5%) |
| history_law_government | History, Law and Government | 9 (0.0%) | 9 (0.1%) |
| media_entertainment | Media and Entertainment | 18 (0.0%) | 16 (0.2%) |
| nature_environment | Nature and Environment | 46 (0.0%) | 38 (0.6%) |
| news_current_affairs | News and Current Affairs | 6 (0.0%) | 6 (0.1%) |
| technology_robotics | Technology and Robotics | 15 (0.0%) | 14 (0.2%) |
| language_fundamentals | Language Fundamentals | 9 (0.0%) | 9 (0.1%) |
Filter out clips with /ŗ/
Filter out clips with characters not in the Latvian alphabet
Each row of a tsv file represents a single audio clip, and contains the following information:
client_id - hashed UUID of a given user
path - relative path of the audio file
sentence - the sentence to be read aloud
sentence_id - unique identifier for the sentence
sentence_domain - domain classification(s) of the sentence
up_votes - number of people who said audio matches the text
down_votes - number of people who said audio does not match text
age - age of the speaker1
gender - gender of the speaker1
accents - accents of the speaker1
variant - variant of the language1
locale - locale code of the language
segment - if sentence belongs to a custom dataset segment, it will be listed here
validated_sentences.tsvThe validated_sentences.tsv file contains one row per validated sentence in the text corpus:
sentence_id - unique identifier for the sentence
sentence - the sentence text
variant - the variant of the language
sentence_domain - the domain(s) the sentence belongs to
source - the source the sentence was collected from
is_used - whether the sentence is still in circulation for recording
clips_count - number of clips recorded for this sentence
unvalidated_sentences.tsvThe unvalidated_sentences.tsv file contains one row per unvalidated sentence in the text corpus:
sentence_id - unique identifier for the sentence
sentence - the sentence text
variant - the variant of the language
sentence_domain - the domain(s) the sentence belongs to
source - the source the sentence was collected from
up_votes - number of upvotes the sentence received
down_votes - number of downvotes the sentence received
status - current status of the sentence (pending or rejected)
Raivis Dejus
This dataset is released under the Creative Commons Zero (CC-0) licence. By downloading this data you agree to not determine the identity of speakers in the dataset.
For a full list of age, gender, and accent options, see the demographics spec. These will only be reported if the speaker opted in to provide that information. ↩ ↩2 ↩3 ↩4