Datasets:

hypaai
/

Hypa-LibreSpeech

text stringlengths 4 565	src_lang stringclasses 8 values	tgt_lang stringclasses 8 values	duration_seconds float64 1.3 22.5	mode stringclasses 1 value	speaker stringclasses 1 value
ook in de allerslechtste dat meestal mijn geval is zitten wil dat ik mij meermalen alleruitmuntendst op een stoomboot heb vermaakt onder anderen ook door al mijn reisgenooten uit te teekenen	dutch	dutch	11.12	transcribe	random
ha de ter noticias minhas mas adeus meu anjo vou-me embora estou com muita pressa é-me indispensavel encontrar em casa a apulina panfilovna para lhe contar o caso	portuguese	portuguese	10.06	transcribe	random
six mois se passèrent encore et l'année d'après charles fut définitivement envoyé au collège de rouen où son père l'amena lui-même vers la fin d'octobre à l'époque de la foire	french	french	13.39	transcribe	random
in the morning at day light we put about to examine the danger we were in and found we had got embayed in a double reef which will very soon be an island	english	english	11.47	transcribe	random
also sprach sie und stieg empor zu den schönen gemächern nicht allein es gingen mit ihr die übrigen jungfraun	german	german	14.89	transcribe	random
sóio el confesor que aunque conformaba con ellos por probarme según después supe siempre me consolaba y me decía que aunque fuese demonio no ofendiendo yo a dios no me podía hacer nada que ello se me quitaría que lo rogase mucho a dios	spanish	spanish	16.1	transcribe	random
avete visto padron ntoni aggiungeva piedipapera dopo la disgrazia di suo nipote sembra un gufo tale e quale adesso la casa del nespolo fa acqua davvero da tutte le parti come una scarpa rotta e ogni galantuomo bisogna che pensi ai suoi interessi	italian	italian	15.17	transcribe	random
étendu sur le trottoir la poitrine trouée et ramassèrent son chapeau qui s'était échappé de sa main la commune très déçu par l'arrivée d'étrangers qu'il n'attendait pas au lieu de celui qu'il espérait	french	french	10.91	transcribe	random
en nog een ander heer hartelijk zat te lachen toen grimaldi midden in het ballet van het tooneel stapte kwam sir godfrey hem te gemoet zeggende zwaar werk grimaldi zwaar en warm sir godfrey	dutch	dutch	19.63	transcribe	random
de koets wegstuurde en den weg naar het etablissement van juffrouw todgers insloeg hoewel het gezicht de gestalte en de gang van den ouden man	dutch	dutch	14.3	transcribe	random
cinco grandes ciudades a porfía baten los yunques y renuevan las armas la poderosa atina la soberbia tíbur árdea crustumera y la torreada antemna	spanish	spanish	15.37	transcribe	random
et il écrivit sur les tables selon ce qu'il avait écrit la première fois les dix paroles que l'éternel vous avait dites sur la montagne du milieu du feu le jour de la congrégation et l'éternel me les donna	french	french	13.81	transcribe	random
wmieszał się między kupu ących i pędząc bez namysłu ceny zakupił biurko i zwierciadło kazał oba nieść za sobą na tragach i do mnie rzekł chodź pan ze mną	polish	polish	10.32	transcribe	random
and he was afraid of finding no room for his exertions i have spoken of the emigration from the older states but how shall i describe that which takes place from the more recent ones fifty years have scarcely elapsed since that of ohio was founded	english	english	15.065	transcribe	random
esto es atroz dios mío carlos se puso de rodillas junto al lecho y le dijo habla qué has comido respóndeme en nombre del cielo y la miraba con ternura nunca le babia visto mirar asi	spanish	spanish	19.9	transcribe	random
indem ich dieses alles glaubte überfiel mich eine solche angst und todessorge daß ich nicht mehr wußte wo ich bleiben sollte und als die musikanten deren ich bisher noch nicht wahrgenommen noch darzu sich hören ließen	german	german	15.24	transcribe	random
la titti era sempre la titti e ogni qual volta la nominava gli occhi gli ridevano umidi di commozione aveva potuto anche argomentare quanto la amasse dalle notizie che le aveva dato sul linguaggio di lei	italian	italian	13.48	transcribe	random
os rapazes precisam passear grifava ele com a liberdade de mordomo confidente aristarco replicava com a invenção cordata dos gêneros de terceira elasticidade insensível dos orçamentos	portuguese	portuguese	16.04	transcribe	random
die allgemeine schwärmerey die meine erscheinung erregte ging anfangs so weit daß ich sogar einem freunde nicht ohne unbescheidenheit davon sprechen kann	german	german	16	transcribe	random
wanneer was het hoe is het gebeurd terwjjl de uitdrukking van ingespannen aanmacht op zgn voorhoofd terugkeerde scheen hg bewust dat die ook op het hare was	dutch	dutch	16.87	transcribe	random
przepraszam pana panie rzecki wtrącił z pokorą gutmorgen ale poco pan darmo pracuje niechże was milion dyabłów porwie krzyknął pan ignacy wybiegł ze sklepu i starannie zamknął drzwi na klucz	polish	polish	14.22	transcribe	random
for the fancy that i dreamed would serve me no longer i saw i felt	english	english	6.22	transcribe	random
fabeln und erzählungen	german	german	12.9	transcribe	random
truth is kitty you'd better dress in monotones she might wake up to the fact that you're a mighty pretty young woman and suddenly become temperamental she has a husband round the lot somewhere make him think his wife is a lucky woman here's all the dope	english	english	15.855	transcribe	random
y enferma y débil se ocupaba en los trabajos mas duros no habia piedad para ella tenia una ama feroz y un amo venenoso el bodegon de thenardier era como una tela de araña donde cosette estaba cogida y temblaba	spanish	spanish	16.87	transcribe	random
habían entablado animada conversación formando otro corrillo no se olvide el señor condito dijo menegilda que nos prometió traer una noche a su novia	spanish	spanish	13.08	transcribe	random
ao gallo bem podeis descer vos seguramente que agora acabou se de assentar paz universal entre todas as aves e animaes por tanto vinde festejaremos este dia	portuguese	portuguese	10.72	transcribe	random
the disturbing something which his mind had unconsciously built up but the new alan revolted he wanted to carry the thing away with him he wanted it to live and so it went with him uncontaminated by any truths or lies	english	english	14.85	transcribe	random
elle n'allait pas désobéir apparemment se mettre en état de péché mortel pour mourir de mort subite et tomber dans l'enfer hugues d'abord ne comprenait rien peu à peu il démêla la trame obscure les racontars probables l'aventure ébruitée	french	french	19.84	transcribe	random
tegenover hem zat een heer die een bijzonder liefhebber van pruimtabak was en wiens lippen en kin daarvan de aangedroogde blijken vertoonden	dutch	dutch	15.08	transcribe	random
we shall corner our game there i'll warrant for this impudent scarlet pimpernel has had the audacity or the stupidity i hardly know which to adhere to his original plans he has gone to meet de tournay saint just and the other traitors	english	english	15.58	transcribe	random
the point is important because what is called thought consists mainly though i think not wholly of inner speech if professor watson is right as regards inner speech this whole region is transferred from imagination to sensation	english	english	14.76	transcribe	random
insulam se em sucessivos sítios e não revêem nunca os caminhos percorridos condenados ao desconhecido afeiçoam se às paragens ínvias e inteiramente novas alcançam nas abandonam nas prosseguem e não se	portuguese	portuguese	20	transcribe	random
e pensando la scelerata matrigna di mandar ad effetto il suo maligno proponimento seminò per tutto il regno che le due figliuole erano morte una di continova febbre l altra per una postema vicina al cuore eh affocata	italian	italian	19.12	transcribe	random
mais il prenait un ton doux avec les bourgeois son costume comme celui des messagers du second ordre consistait en de bonnes grosses bottes pesantes de clous faites à l'isle-adam et un pantalon de gros velours vert-bouteille et une veste de semblable étoffe	french	french	18.37	transcribe	random
y guando lo hubo orde nado como convenia y haber reeebido los san tos sacramentos fue nuestro señor jesu christo servido de llevarle deste trabajoso mundo y murió en dos dias del mes de diciembre de mil y quinientos y quarenta y siete años	spanish	spanish	16.72	transcribe	random
foi a primeira vez que passou pelas duras provas o animal informe atirou-se a ella por entre uma chuva de faiscas abrasadoras ella porém deitou a correr pelo isthmo a fora como se tivesse perdido a razão	portuguese	portuguese	18.54	transcribe	random
but what he loved and valued above all was the money he had amassed by his labour and by all sorts of devices that money made him the equal of all who had been his superiors	english	english	9.64	transcribe	random
de jonge man zat daar heel huiselijk achter een boezelaar die daar te drogen hing te peinzen terwijl ik hem aanzag dacht ik onwillekeurig aan w	dutch	dutch	15.79	transcribe	random
the eagerness of the penniless children to get into these magic spaces is responsible for an entire crop of petty crimes made more easy because two children are admitted for one nickel at the last performance when the hour is late	english	english	13.985	transcribe	random
da praça principal da vila junto à garganta que conduz à pequena praça cotovelo nos fundos da casa indicada era então a embocadura do riacho da palha que em forma quase circular contornava aquela praça e de inverno constituía uma cinta	portuguese	portuguese	20	transcribe	random
dove per altro così bagnati fino all'osso non avremmo potuto rimanere che peccato le dissi si doveva star qui un'ora almeno a finire la storia incominciata un'ora esclamò doveva durar tanto quella brutta storia	italian	italian	12.63	transcribe	random
ma in vano s affaticava in sparger il fiato perciò che la misera alma era partita di questa vita e se ne era ita all'altra l altro compagno vedendo questo disse oh sciocco tu non hai saputo ben fare lascia un poco fare a me	italian	italian	16.65	transcribe	random
niet op government house gei'nviteerd werd verklaarde dat zijne excellence een tyran en nero bij hem vergeleken een verlicht philanthroop was	dutch	dutch	15.16	transcribe	random
warum edle frau wollen sie sich so oft der bösen luft die hier herrscht aussetzen sollte denn das schicksal mit ihnen so hart sein daß sie zu sterben begehrten	german	german	13.39	transcribe	random
toen zij ging slapen koos rebecca den laatste om van te droomen om vier uur op zulk een stralenden zomerochtend die zelfs	dutch	dutch	16.83	transcribe	random
wept lydia as a rule he is always smart in replying	english	english	3.79	transcribe	random
não se ria nhô mundico não se ria prosseguiu a sogra de manuel que aqui está e bateu no peito quem já andou de quebranto a dar não dá com os ossinhos no gavião e tirando do seio um trancelim com uma enorme figa de chifre encastoada em ouro	portuguese	portuguese	14.67	transcribe	random
and sent him back loaded with the favours i have enumerated in short i employed all my eloquence to persuade him to imitate so good an example and to grant me pardon but it was impossible to move his compassion	english	english	14.175	transcribe	random
à porta da igreja a sra tomásia velha devota que o adora vem ao encontro dele padre joão aqui está um regalo que lhe quero oferecer para o seu almocinho de hoje	portuguese	portuguese	12.72	transcribe	random
ma poco dopo all'improvviso non potendo interessarsi di quelle vuote chiacchiere era riassalito da quell'immagine e si sentiva schernito da quella gente	italian	italian	12.05	transcribe	random
o dedo do douto historiador ia-me apontando todos os lugares religiosos cujos nomes sonoros caem na alma com uma solenidade de profecia ou com um fragor de batalha esdrelon endor sulém tabor	portuguese	portuguese	16.1	transcribe	random
y el pedro barba y los demas que consigo traian preguntan por el señor pánfilo de narvaez y cómo le va con cortés y responden que muy bien que cortés anda huyendo y alzado con veinte de sus compañeros que narvaez está muy próspero y rico y que la tierra es muy buena	spanish	spanish	19.2	transcribe	random
nagle zabrzmiał głos ochrypły świszczący chropawy co to co to wtem zoryentowałem się był to mój śmiech	polish	polish	16.56	transcribe	random
zijn tobben waarom een almachtig god zooveel onrecht duldt de botsingen van zijn ontwakend idealisme met het bestaande thuis op school in den handel dat alles wordt in fijn gevoelde humoristische tafreeltjes geschetst	dutch	dutch	13.65	transcribe	random
częściej bierze na rachunek więc bierze tu wokulski odetchnął jakże on stoi zdaje się że to skończony bankrut i bodaj że w tym roku zlicytują mu nareszcie kamienicę wokulski pochylił się nad kanapą i zaczął bawić się z irem proszę cię a panna łęcka nie wyszła zamąż	polish	polish	18.91	transcribe	random
that fighting would probably go on for a long time yet and that things being so it was quite likely he might be in command of a regiment in a couple of years time as he looked at the matter in this way	english	english	11.75	transcribe	random
soltanto ma proprio appena egli poteva ancora tentare di muovere una mano la sinistra dopo essersela guardata a lungo con quegli occhi quasi a infonderle il movimento	italian	italian	14.21	transcribe	random
le gouvernement de la défense nationale considérant qu'à la suite d'excitations criminelles dont certains clubs ont été les foyers la guerre civile a été engagée par quelques agitateurs désavoués par la population tout entière	french	french	16.06	transcribe	random
dazu also herzen zergliedert im dunkel der seelen gewühlt mit richterkunst und pathos tat und untat auf ihr menschlich maß geprüft	german	german	10.46	transcribe	random
good luck is rather particular who she rides with and mostly prefers those who have got common sense and a good heart at least that is my experience governor gray turned round again to his newspaper	english	english	14.27	transcribe	random
denn die bauern tschitschikows werden zwei mächtige feinde vor sich haben der erste feind ist die nähe der kleinrussischen gouvernements wo bekanntlich freier branntweinverkauf besteht ich versichere sie in zwei wochen werden sie dem suff erliegen	german	german	17.73	transcribe	random
y sin siquiera ladrar por obedecer a su amo seguiré tu consejo hernando repuso el caballero lanzando un suspiro le seguiré y con la ayuda de dios y de mi buen caballo estaremos al alba fuera de madrid recógete pues hernando y descansa	spanish	spanish	18.81	transcribe	random
e la stella che vede ne parla al cielo infinito ah in vano muore sfugge alla morta pupilla già il bimbo che geme al suo piede	italian	italian	17.51	transcribe	random
en maakte eene diepe buiging zij nam de eereplaats in en gaf mijnen broeder een wenk dat ook hij zich weêr zou nederzetten terwijl zij op lagchenden toon tot hem zeide	dutch	dutch	13.67	transcribe	random
ein vergebliches bemühen kein gedanke kam ihm in den kopf bald versuchte er an nichts zu denken vergebliche mühe bruchstücke von gedanken enden und zipfel von gedanken kamen ihm von allen seiten in den sinn	german	german	15.19	transcribe	random
der bewoners van bet huis de kinderen waren wel is waar zindelijk maar hunne kleederen waren oud en versleten men had geen huurder voor de bovenkamer kunnen krijgen op welker opbrengst men gerekend had om de'huur te kunnen betalen	dutch	dutch	18.02	transcribe	random
er sagte ihr scherzend sie sei doch jung und müsse sich unterhalten und zerstreuen und dürfe sich von so einem alten langweiligen menschen wie er nicht anöden lassen	german	german	11.94	transcribe	random
l'un sur les signatures des peintres célèbres et sur les moyens de reconnaître l'authenticité de leurs oeuvres l'autre sur l'art de l'encadrement à la suite de quoi il avait été nommé officier d'académie	french	french	11.65	transcribe	random
nawet gdyśmy się dopytywali widząc czasem że mizernie wygląda wstrząsała tylko głową i mówiła z uśmiechem nic mi nie jest albo to przejdzie nie umrę jeszcze bo jeszcze jestem tomowi potrzebna	polish	polish	14.17	transcribe	random
les dames mirent un certain empressement à quitter le musée le tomahawk de théodule les inquiétait en allant déjeuner à l'hôtel de la bosse de biso i philippe regardait toujours les enseignes tout à coup il se frappa le front	french	french	14.15	transcribe	random
nawet kapłanów ale tutmozis w tych drwiących słowach odczuł groźbę książę bardzo kochał wiernego jak pies patroklesa mógł zapomnieć wiele własnych krzywd ale jego śmierci nie przebaczy nigdy	polish	polish	14.28	transcribe	random
iedereen in de kamer droeg met echt republikeinschen vrijheidszin dat teeken der mannelijke oppermacht op het hoofd hetzij van vilt of palmbladeren oud en smerig of glimmend nieuw	dutch	dutch	17.2	transcribe	random
e gridare un viva sonoro davanti all'enorme splendida tumultuosa temeraria tela del makart tutta irradiata dal viso bianco di carlo v su cui brilla un pensiero vasto come il suo regno	italian	italian	16.28	transcribe	random
a jednak ten mraczewski jest infamis myślał jak można mówić takie rzeczy w sklepie za parę dni otrzymam bilecik a potem schadzka ha sama sobie winna nie trzeba kokietować błaznów zresztą wszystko mi jedno	polish	polish	14.84	transcribe	random
sobrinho apareceu aborrecido a sobrinha triste o diálogo foi mastigado como o almoço no fim deste recebeu estácio uma carta de eugênia era uma tagarelice meio frívola meio sentimental mistura de risos e suspiros sem objeto definido a não ser pedir-lhe que escrevesse se não pudesse	portuguese	portuguese	19.82	transcribe	random
t gebeurde spoediger dan zijzelf had verondersteld het was doodstil in het huis der bornes de doodse stilte die in het binnenland van java heerst tussen het derde en vierde uur na middernacht als al wat leeft schijnt te slapen	dutch	dutch	16.37	transcribe	random
weg wodurch die kassen und magazine eine menge remontepferde und sämtliche caesar gestellte geiseln den insurgenten in die hände fielen	german	german	16.21	transcribe	random
di colpo si guardarono si tesero le mani contemporaneamente stringendosele si erano fermati per un secondo addio disse il signor roberto addio rispose manlio	italian	italian	12.83	transcribe	random
me pareció oir voces apagadas de gente que vagaba por el huerto el perro había enmudecido las voces se desvanecían de nuevo quedó todo en silencio y en medio del silencio oí el galope de un caballo que se alejaba	spanish	spanish	16.68	transcribe	random
lo sguardo intento tra il vasto arco cigliare così svelta di forme nella guaina rosa la nera chioma ondosa chiusa nel casco enorme ed in quellurna appesa con quella fitta rete dormono cento quete crisalidi in attesa	italian	italian	16.99	transcribe	random
parlò con amore di suo padre era ingegnere militare nell'esercito austriaco era assai colto sapeva lo spagnuolo l'inglese il francese il tedesco pubblicò vari scritti scientifici che lo zola conserva	italian	italian	14.68	transcribe	random
a ce mot arrêtons-nous et plaçons ici pour les ignorants une explication due à un étymologiste très-distingué qui a désiré garder l'anonyme bourguignon est le nom populaire et symbolique donné depuis le règne de charles vi	french	french	15.25	transcribe	random
mości panie mówił cudzoziemiec zaręczam panu że potrafię tu wejść moja narzeczona naznacza sobie w tym domu schadzki miłosne z jakimś mieszczaninem z kadyxu wiem o tem z pewnością	polish	polish	12.33	transcribe	random
his favorite thesis included the origin of mammalian life and of man himself in southernmost south america with as incidents the belief that the mammalian bearing strata of south america	english	english	13.46	transcribe	random
ya se conoce bien no le ocultaré á usted ahora que como tengo una recta conciencia me hallaré intranquilo mientras no remedie el mal que involuntariamente he causado á la inocente niña que usted ama	spanish	spanish	14.2	transcribe	random
vir o mesmo respondeu estácio ou ainda melhor melhor decerto porque dois anos mais modificam o homem estácio fez aqui um panegírico do amigo intercalado com observações da tia e ouvido silenciosamente pela irmã	portuguese	portuguese	14.36	transcribe	random
and competent to grapple with anything or anybody there was the queer old gentleman who had crossed eleven times before and had advice and experience to spare for any one who would listen to them and the other gentleman not so old but even more queer	english	english	14.825	transcribe	random
ik zou het doen tom als er iemand vr mij was wanneer ik bid maar het is alles in de lucht gesproken als ik het doe maar kom aan tom bidt gij en leer mij het te doen	dutch	dutch	17.09	transcribe	random
the younger professors the illiterate athletes like langueduc think he's getting eccentric but they just say good old burne has got some queer ideas in his head and pass on the pharisee class gee they ridicule him unmercifully	english	english	15.05	transcribe	random
ce n'était pas évidemment en elle-même une terminaison bien extraordinaire mais l'immobilité qui l'avait précédée la faisait se détacher avec la netteté cristalline l'imprévu quasi malicieux de ces phrases par lesquelles le piano	french	french	16.56	transcribe	random
e che tutta l'america ne dovesse esser coperta guardava attentamente i nomi delle vie dei nomi strani che stentava a leggere a ogni nuova via si sentiva battere il cuore pensando che fosse la sua guardava tutte le donne con l'idea di incontrare sua madre	italian	italian	18.98	transcribe	random
ces sages opérations méditées entre le docteur et le juge de paix furent accomplies dans le plus profond secret à la faveur des troubles politiques	french	french	11.8	transcribe	random
laten wy dan onze zorgvuldigheid besteden voor hunne kinderen en aan die zo veel goed doen als wy kunnen heeft de voorzienigbetje wolff proeve over de opvoeding heid u met overvloed bedeelt ô besteed daar van een gedeelte ten besten hunner kinderen welingerichte scholen zullen hier denkelyk het best aan het oogmerk	dutch	dutch	18.62	transcribe	random
poszedł nad staw obleciał park wokoło jakby chcąc zgubić w drodze złe przeczucia napróżno uczepiła go się myśl że panna izabela może wyjechać tłumił ją i przytłumił o tyle że już nie rysowała mu się jasno tylko gdzieś na dnie serca drażniła go nieznacznie	polish	polish	17.47	transcribe	random
en sortant de la séance académique mme ponto conduisit hélène aux bureaux de y époque m hector piquefol était là présidant à la rédaction du numéro du soir	french	french	11.77	transcribe	random
pouco tempo depois d isabel thereza de souza quintella era tambem com ordem de captura conduzida á quinta de sua mãi nos arrabaldes de lisboa levava o filho nos braços	portuguese	portuguese	12.33	transcribe	random
uma vez muito entusiasmado o ilustre mestre mostrou nos o cruzeiro do sul pouco depois cochichando com o que sabíamos de pontos cardeais descobrimos que a janela fazia frente para o norte não atinamos aristarco reconheceu o descuido não quis desdizer se	portuguese	portuguese	16.38	transcribe	random
okolica wydała mi się czarowną pola połyskiwały najświetniejszemi barwami spostrzegłam także w oczach mego brata pewien zapał odmienny od tego jaki w nim przedtem gorzał do nauk	polish	polish	12.29	transcribe	random
la bocca del burattino pareva inchiodata e ribadita allora lassassino più piccolo di statura cavato fuori un coltellaccio provò a conficcarglielo a guisa di leva e di scalpello fra le labbra ma pinocchio lesto come un lampo gli azzannò la mano coi denti	italian	italian	16.01	transcribe	random

End of preview. Expand in Data Studio

hypa_libreSpeech_data_ASR_Banner

Hypa-LibreSpeech

Hypa-LibreSpeech is a large-scale, multilingual speech dataset curated by Hypa AI for training and evaluating Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) systems across 8 European languages. It contains 200,000 high-quality audio–text pairs derived from open-domain audiobook recordings originally sourced from the LibriVox project.

This dataset builds directly upon two foundational open-source corpora:

openslr/librispeech_asr — the original English LibriSpeech corpus by Panayotov et al. (2015)
facebook/multilingual_librispeech — the Multilingual LibriSpeech (MLS) corpus by Pratap et al. (2020), covering 8 languages derived from LibriVox audiobooks

Hypa-LibreSpeech extends and repackages these sources into a unified, streamable, Whisper-ready format designed for modern multilingual speech model training.

This dataset is part of the broader Hypa AI open data initiative, which aims to democratize access to high-quality speech data for AI researchers and developers worldwide.

Dataset Summary

Hypa-LibreSpeech is an open-source multilingual speech dataset derived from the LibreSpeech corpus and curated to support the development of robust speech and language technologies across multiple languages. The dataset contains approximately 200,000 high-quality audio–text pairs spanning 8 languages, providing a diverse collection of speech recordings from multiple speakers, accents, and speaking styles.

The dataset includes aligned speech recordings and text transcriptions, making it suitable for a wide range of speech processing and language modeling tasks. Audio samples range from 1.3 to 22.5 seconds in duration, while transcription lengths vary from 4 to 565 characters, capturing both short utterances and longer spoken passages. To support different deployment scenarios, audio is provided in both FLAC (lossless) and OPUS (compressed) formats.

This release provides:

Multilingual Speech–Text Pairs: Paired audio recordings and transcriptions across 8 languages. Multiple Audio Formats: Speech data available in FLAC and OPUS formats for training and deployment flexibility. Diverse Speaker Coverage: Recordings from a broad pool of speakers with varying accents and speaking characteristics. Unified Transcription Task: Standardized speech-to-text samples designed for multilingual automatic speech recognition research.

The dataset is designed to support a variety of downstream tasks, including:

Automatic Speech Recognition (ASR): Fine-tuning and evaluating multilingual ASR models such as Whisper, Wav2Vec2, and MMS. Text-to-Speech (TTS): Training multilingual and cross-lingual speech synthesis systems such as XTTS, VITS, and Orpheus. Speech Translation Research: Building and evaluating multilingual speech processing pipelines. Cross-Lingual Transfer Learning: Investigating knowledge transfer between high-resource and low-resource languages. Speech Representation Learning: Pretraining and benchmarking speech foundation models. Language Identification and Classification: Developing systems for spoken language recognition and multilingual speech understanding.

By providing a large-scale multilingual collection of aligned speech and text data, Hypa-LibreSpeech aims to facilitate research and development of inclusive speech technologies that generalize effectively across languages and speaker populations.

Supported Tasks and Leaderboards

Task	Description
`automatic-speech-recognition`	Transcribe spoken audio to text in the source language
`text-to-speech`	Use paired text–audio for voice synthesis training
`audio-classification`	Language identification from audio

Languages

The dataset covers 8 European languages, matching the language scope of the upstream MLS corpus:

Language	ISO Code	Script
English	`en`	Latin
French	`fr`	Latin
German	`de`	Latin
Dutch	`nl`	Latin
Spanish	`es`	Latin
Italian	`it`	Latin
Portuguese	`pt`	Latin
Polish	`pl`	Latin

Dataset Structure

Data Instances

Each instance in the dataset represents a single utterance and contains the following:

{
  "audio": {
    "path": "audio/train/0001.opus",
    "array": [...],          # decoded audio waveform
    "sampling_rate": 16000
  },
  "text": "ook in de allerslechtste dat meestal mijn geval is...",
  "src_lang": "dutch",
  "tgt_lang": "dutch",
  "duration_seconds": 11.12,
  "mode": "transcribe",
  "speaker": "random"
}

Data Fields

Field	Type	Description
`audio`	`Audio`	Audio object containing the waveform array, path, and sampling rate (16,000 Hz)
`text`	`string`	Ground-truth text transcription of the audio clip
`src_lang`	`string`	Source language of the audio (e.g., `"dutch"`, `"english"`)
`tgt_lang`	`string`	Target language of the transcription (same as `src_lang` for transcription tasks)
`duration_seconds`	`float`	Duration of the audio clip in seconds
`mode`	`string`	Task mode — currently `"transcribe"` for all instances
`speaker`	`string`	Speaker label — currently `"random"`, reflecting the diverse volunteer speaker pool

Data Splits

Split	Num. Examples
`train`	200,000

The dataset ships as a single train split. Users are encouraged to create their own validation and test splits as needed. For reference, the upstream MLS corpus provides standardized dev and test splits per language that can be used for evaluation.

Dataset Creation

Source Data

Hypa-LibreSpeech is derived from two primary upstream corpora:

1. LibriSpeech ASR Corpus

Panayotov, V., Chen, G., Povey, D., & Khudanpur, S. (2015). LibriSpeech: An ASR corpus based on public domain audio books. ICASSP.

Available on Hugging Face as openslr/librispeech_asr.

LibriSpeech is a corpus of approximately 1,000 hours of 16kHz read English speech, prepared by Vassil Panayotov with the assistance of Daniel Povey. It is derived from read audiobooks from the LibriVox project, and has been carefully segmented and aligned.

2. Multilingual LibriSpeech (MLS)

Pratap, V., Xu, Q., Sriram, A., Synnaeve, G., & Collobert, R. (2020). MLS: A Large-Scale Multilingual Dataset for Speech Research. ArXiv:2012.03411.

Available on Hugging Face as facebook/multilingual_librispeech.

MLS is a large multilingual corpus derived from read audiobooks from LibriVox, covering 8 languages: English, German, Dutch, Spanish, French, Italian, Portuguese, and Polish. It includes approximately 44,500 hours of English and a total of approximately 6,000 hours for the other 7 languages, making it one of the largest publicly available multilingual speech datasets.

3. LibriVox (Original Audio Source)

All audio ultimately originates from LibriVox, a volunteer-driven project that produces free public domain audiobook recordings. LibriVox releases all recordings under the CC0 1.0 Universal (Public Domain) license.

Curation Rationale

While MLS and LibriSpeech are foundational resources, they are not always straightforward to use directly for modern training pipelines — particularly for fine-tuning instruction-following speech models like Whisper. Hypa-LibreSpeech addresses this by:

Unifying schema: Providing a single consistent schema across all 8 languages with src_lang, tgt_lang, mode, and duration_seconds fields
Filtering for quality: Removing segments with transcription alignment issues, excessively short clips (< 1.3s), or very long clips (> 22.5s)
Optimizing for streaming: Encoding audio in OPUS format for efficient streaming alongside lossless FLAC archives
Repackaging for modern use: Structuring data to be immediately usable with HuggingFace datasets, transformers, and common training frameworks

Processing Pipeline

The dataset was produced through the following pipeline:

Source ingestion — Audio and aligned text were ingested from the MLS and LibriSpeech corpora via the Hugging Face Datasets library.
Segmentation validation — Segment boundaries were validated against the original LibriVox source timestamps. Segments with misalignments were discarded.
Duration filtering — Clips shorter than 1.3s or longer than 22.5s were removed to ensure training stability.
Text normalization — Transcriptions were lowercased and lightly normalized (removing chapter headers, annotations, and non-speech markers).
Audio re-encoding — Segments were re-encoded to 16 kHz mono in both FLAC (lossless) and OPUS (compressed, ~50% size reduction).
Language tagging — Each segment was tagged with src_lang and tgt_lang derived from the originating corpus language metadata.
Schema standardization — All fields were aligned to the unified Hypa-LibreSpeech schema and exported to Parquet for efficient loading.

Statistics

Metric	Value
Total Examples	200,000
Total Languages	8
Min Duration	1.3 seconds
Max Duration	22.5 seconds
Estimated Total Audio	~600 hours
Audio Formats	FLAC, OPUS
Sampling Rate	16,000 Hz
Text Min Length	4 characters
Text Max Length	565 characters
Speaker Pool	Mixed (volunteer readers via LibriVox)
Task Mode	`transcribe`

Approximate Language Distribution

Language	Approx. Examples	Source
English	~25,000	LibriSpeech + MLS
French	~25,000	MLS
German	~25,000	MLS
Dutch	~25,000	MLS
Spanish	~25,000	MLS
Italian	~25,000	MLS
Portuguese	~25,000	MLS
Polish	~25,000	MLS

For reference, the full upstream MLS corpus contains the following training hours per language:

Language	MLS Train Hours	MLS Train Samples
English	44,659	—
German	1,967	469,942
Dutch	1,554	374,287
French	1,077	258,213
Spanish	918	220,701
Italian	247	59,623
Portuguese	161	37,533
Polish	104	25,043

Hypa-LibreSpeech draws a balanced 200k-sample subset from these sources.

Usage

Loading the Dataset

from datasets import load_dataset

# Load the full dataset
dataset = load_dataset("hypaai/Hypa-LibreSpeech")

# View a sample
print(dataset["train"][0])

Filtering by Language

from datasets import load_dataset

dataset = load_dataset("hypaai/Hypa-LibreSpeech", split="train")

# Filter for French examples only
french_data = dataset.filter(lambda x: x["src_lang"] == "french")
print(f"French examples: {len(french_data)}")

Fine-tuning Whisper

from datasets import load_dataset
from transformers import WhisperForConditionalGeneration, WhisperProcessor

# Load dataset
dataset = load_dataset("hypaai/Hypa-LibreSpeech", split="train")

# Load Whisper processor and model
processor = WhisperProcessor.from_pretrained("openai/whisper-small")
model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-small")

# Prepare a single sample
sample = dataset[0]
inputs = processor(
    sample["audio"]["array"],
    sampling_rate=sample["audio"]["sampling_rate"],
    return_tensors="pt"
)

# Generate transcription
predicted_ids = model.generate(inputs["input_features"])
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
print(transcription)

Streaming for Large-Scale Training

from datasets import load_dataset

# Stream without downloading the full dataset
dataset = load_dataset("hypaai/Hypa-LibreSpeech", split="train", streaming=True)

for sample in dataset.take(5):
    print(sample["text"], "|", sample["src_lang"], "|", sample["duration_seconds"], "s")

Using with PyTorch DataLoader

from datasets import load_dataset
from torch.utils.data import DataLoader

dataset = load_dataset("hypaai/Hypa-LibreSpeech", split="train", streaming=True)
dataloader = DataLoader(dataset, batch_size=16)

Considerations for Using the Data

Social Impact

Hypa-LibreSpeech contributes to the democratization of multilingual speech AI by making a large, well-structured, open-access speech corpus freely available. By lowering the barrier to building ASR and TTS systems across 8 European languages, it supports researchers, startups, and developers who may not have the resources to curate their own large-scale training data.

Biases

Speaker demographics: Audio originates from LibriVox volunteer readers, who skew toward native or near-native speakers and do not uniformly represent all regional accents, dialects, or age groups. Refer to the MLS paper (Pratap et al., 2020) for detailed speaker gender statistics per language.
Domain bias: All text is from literary/audiobook sources (19th–early 20th century literature). This differs significantly from conversational, spontaneous, or technical domain speech. Models trained exclusively on this data may underperform on informal or domain-specific audio.
Language imbalance: The upstream MLS corpus has significant imbalance in total training hours across languages (English: ~44k hours vs. Polish: ~104 hours). Hypa-LibreSpeech rebalances this with a ~25k sample cap per language, but downstream model performance may still vary.
Text style: Transcriptions reflect the literary register of the source texts, including complex sentence structures and archaic vocabulary in some languages.

Limitations

The dataset contains a single train split. Users must create their own validation and test sets, or use the standardized splits from the upstream facebook/multilingual_librispeech dataset for evaluation.
The speaker field is set to "random" for all entries. Speaker-level diarization or attribution is not included in this release.
Audio quality varies naturally due to the volunteer recording nature of LibriVox — some recordings contain minor background noise, room echo, or microphone variation.
The dataset does not include timestamps at the word or phoneme level.

Licensing Information

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license, consistent with the license of the upstream MLS corpus.

The underlying audio recordings are sourced from LibriVox, released into the public domain under the CC0 1.0 Universal license.

You are free to use, share, and adapt this dataset for any purpose, including commercial use, provided appropriate attribution is given to Hypa AI and the upstream sources.

Citation Information

If you use Hypa-LibreSpeech in your research or projects, please cite this dataset and the upstream corpora it is derived from:

Cite Hypa-LibreSpeech

@dataset{hypaai2025librespeech,
  title        = {Hypa-LibreSpeech: A Multilingual Audiobook Speech Dataset},
  author       = {Hypa AI},
  year         = {2025},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/hypaai/Hypa-LibreSpeech},
  license      = {CC BY 4.0},
  note         = {200,000 audio-text pairs across 8 European languages, derived from LibriSpeech and Multilingual LibriSpeech (MLS).}
}

Cite Multilingual LibriSpeech (MLS)

@article{Pratap2020MLSAL,
  title   = {MLS: A Large-Scale Multilingual Dataset for Speech Research},
  author  = {Vineel Pratap and Qiantong Xu and Anuroop Sriram and Gabriel Synnaeve and Ronan Collobert},
  journal = {ArXiv},
  year    = {2020},
  volume  = {abs/2012.03411},
  url     = {https://arxiv.org/abs/2012.03411}
}

Cite LibriSpeech

@inproceedings{panayotov2015librispeech,
  title     = {LibriSpeech: An ASR corpus based on public domain audio books},
  author    = {Panayotov, Vassil and Chen, Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
  booktitle = {2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages     = {5206--5210},
  year      = {2015},
  organization = {IEEE}
}