Datasets:

suleiman2003
/

W_hausa_v7

text stringlengths 3 504	speaker_id int64 1 55	gender stringclasses 2 values	duration float32 1 29.8
Kwata sau ashirin= ya zama biyar.	1	male	3.808
Bakin kare ne yana wasa da kwallo mai launukan fari da ja da shudi.	1	male	6.368
Yau za a kasance cikin sauyin yanayi a wunin gaba daya.	1	male	5.344
Nan haske da zafin rana za a sha a makon gaba daya. Sai a lura da cunkosa a daki don kada a kamu da wasu cututtukan zafi.	1	male	11.712
Litattafai guda shida wani kan wani. Na kasan mai launin ja na sama da shi mai launin shudi, na sama da shi guda biyu masu launin kore sai na sama da su mai ja sannan kuma na Samansa mai launin shudi.	1	male	15.808
Da sassafe za a tashi da ruwan sama. Amma daga karfe hudu na yamma rana zata hudo gari ya washe har zuwa faduwarta. Haka za a yi ta yi a wannan ranar.	1	male	11.648
Nan tagar wani gini ne mai launin ruwan kasa ita kuma fara ce, ga alamun duhun cikin dakin can an dage labule, sannan ga wata 'yar tukunyar shuka furanni da kwalliyar gida kan alkukin tagar, har ma da wata shuka ta tsiro a ciki.	1	male	22.048
Ka fara tukin da nufan yankin Arewa a titin Dizengoff	1	male	6.08
Iliya mutum me kwazo sosai	1	male	3.36
Idan ka je Ƙasar Botswana za ka ga inda aka shuka bishiyoyi masu ban sha'awa.	1	male	8.352
Duniyar wata nake son zuwa don ganin abin mamaki	1	male	4.416
Lallai malaman jinyar nan sun kokarta wajen kubutar da jaririn.	1	male	5.952
Hawwa ta je gona jiya da safe.	1	male	2.08
Janar Agui Ironsi shi ne ya zama shugaban ƙasar Nijeriya na farko bayan kifar da gwamnatin su Sardauna.	1	male	9.632
Janar Babangida yana ɗaya daga cikin shugabannin ƙasar Nijeriya da ya yi aikin a-zo-a-gani sai da ya shafe shekaru takwas kan Mulki.	1	male	11.712
Daga nan sai ka bi hanyar bulla zuwa babban titi na biyu, wato ta jikin rafi. Ci gaba da bin babban titin ta bangaren Arewa har na tsawon kilomita casa'in. Sai ka ɗauki hanyar Haifa	1	male	14.24
A jihar Neja akwai babbar madatsar ruwa na Kayinji da ke samar da wuta Nijeriya	1	male	9.728
Lugard hall nan inda Gwamna-janaral ya zauna a Kaduna lokacin Turawan mulkin mallaka. Shi ya sa ake kiran wurin da Lugard hall. Kuma a yanzu nan ne zauren Majalisar dokokin Jihar Kaduna.	1	male	15.616
Suna tafiya kasuwa ranar Larabar aka kama su.	1	male	4.832
Yana tsaye daidai wirin shudiyar motar can.	1	male	4.992
Lokacin Kirsimati muna taya mabiya addinin Kirista murna.	1	male	6.208
Busasshen reshen bishiya maciji ya nannade Shi mai launin ruwan kasa da zane-zane.	1	male	8
Yau za a wuni da tsananin rana.	1	male	3.264
Alamomin wasu zane masu launukan kore mai duhu da mai haske da ja da rawaya da kuma ruwan kasa.	1	male	8.768
Da'ira mai launin shudin mai haske da zanen farin tauraro a tsakiya	1	male	6.88
Rana ta fito ta haske wuri, sai ga wani Jan doki ya keto a guje mai farin gashi a doron wuyansa da wutsiyarsa.	1	male	10.336
Yau za a wuni cikin buji da iska. A sanya takunkumin rufe baki da hanci.	1	male	8.192
A nan fa da alamun za a sha sanyi har a zama kamar kankara. A nemi manyan bargo da rigunan sanyi da mayafi masu kauri kada sanyi ya daskarar da mitane.	1	male	12.608
Tsuntsaye guda uku a sama suna shawagi masu launin rawaya da baki da fari da baki a cikin idanusu.Haka ma bakinsu baki ne da fukafukai masu launin kore mai haske. Sai koriyar tsirrai a kasa da jan fure.	1	male	18.656
Sai ka kai nisan Kilomita dari takwas da hamsin	1	male	4.896
Gambo ya koma kauye da zama.	1	male	2.784
Wai shin da gaske ne Kogin Kwara da ya taho ya rabu biyu nan aka samu kogin Neja da na Binuwai ? kwarai kuwa	1	male	10.336
Kaura idan aka nika aka mulmula tuwo da gari, ai abin ba a magana.	1	male	6.368
Miliyan daya sau day a raba gida biyar ta zama dubu dari biyu kenan.	1	male	5.728
A garin Zuma na hanyar Kaduna zuwa Abuja akwai wani katon dutse da aka hotonsa a bayan naira dari daya	1	male	8.384
A jihar Bauchi akwai wurin shakatawa na Yankari mai dauke da namun daji	1	male	6.816
a-Kowanne mutum yana da 'yancin zama dan wata kasa b- Babu wanda za a kwace wa 'yancinsa na zama dan kasa ba tare da kwakkwaran dalili ba, haka kuma ba za a hana shi damar sauya kasa ba idan yana so.	1	male	16.063999
Ranar Talata mun isa gida lafiya.	1	male	3.488
Na yi masa gargadin ya daina taɓa min farar motar nan.	1	male	5.504
Mutane uku sun mutu ranar Litini saboda sun kamu da cutar zazzabin bera	1	male	5.824
Taswirar mutum ne yana buga kwallo.	1	male	3.456
Rabi sau dari= shi ne hamsin.	1	male	4.416
Algashin da'ira hade da alwatika mai launin toka	1	male	5.76
Wannan saniya ce mai launin fari da baki.	1	male	4
Bishiyar ayaba koriya shar da nunannu 'ya'yanta guda uku.	1	male	5.248
Shudin da'ira hade da alwatika shudi a kansa.	1	male	4.16
Wannan kogi ne da dare lokacin hasken farin wata.	1	male	5.472
An bugo kwallo ta shiga raga ga gajimare a sararin samaniya	1	male	6.08
A nan tumatiri ne guda biyu nunannu da karas wanda ya yi kwarai.	1	male	7.712
Bayan asubahi ke nan, a yayin da rana ta hudo tana fitowa ta bayan tsauni wuri ya yi haske.	1	male	7.232
Murabba'i uku masu launin ruwan kasa a jere kusa da da'ira guda biyu masu launin shudi.	1	male	7.648
uku bisa biyar sau dari da shirin ya zama casa'in	1	male	4.864
Za a samu kadawar iska kadan-kadan.	1	male	3.936
Wannan keken hawa ne baki lafiyayye. Sai a hau a ruga a guje.	1	male	6.72
Koren ganye da fure mai launin algashi ga wani a tsakiya mai launin rawaya.	1	male	7.744
Dari daya sau biyar a raba gida hudu ya zama ashirin da biyar ke nan	1	male	4.96
Shida a haɗa da bakwai= ya zama goma sha uku.	1	male	4.544
Lemon zaki guda nunannu da rabi guda daya da aka yanko daga jikin wata.	1	male	7.552
Daya a haɗa da biyu = ya zama uku.	1	male	3.936
Da'ira mai launin algashi da zane murabba'i uku masu launin ruwan kasa a tsakiya	1	male	7.136
Alwatika biyu bakake sun sanya da'ira mai launin ruwan kasa a tsakiya.	1	male	6.272
Butar dafa shayi mai launin baki da fari. Za ta yi amfani kwarai ainun a wannan lokacin da ake fama da sayi	1	male	9.152
Ma'aunin yanayi takwas, za a sha sanyi har da kankara.	1	male	5.472
Biyu bisa biyar sau tamanin ya zama tamanin da biyu	1	male	5.312
Nunannen aful guda daya da rabi mai launin kore.	1	male	5.952
Saniya mai launin baki da fari a karkashin bishiya tana cin ciyawa ga rana ta kwallaro	1	male	8.352
Daya bisa uku sau sittin= ya zama ashirin.	1	male	4.704
Nan ya nuna ma'aunin yanayi ashirin. Za a samu hasken rana.	1	male	5.408
Daji da bishiyoyi koraye shar da tsauni kewaye da ciyawa mai ban sha'awa.	1	male	7.968
Yabanya koriya shar a gona da wani shudin tsauni can a gefen gonar. Ga gajimare a sararin sama.	1	male	9.984
Murabba'i mai launin kore mai haske da da'ira guda hudu jikinsa ta gaba masu launin kore mai duhu.	1	male	9.632
Daya bisa biyar sau (ashirin cikin Baka a haɗa da talatin a haɗa arba'in)= ya zama goma sha takwas.	1	male	8.832
Goma sha biyu sau hudu= ya zama arba'in da takwas..	1	male	5.056
Ana iya wuni yau da sanyi da iska har mutane su ji ba su son fita.	1	male	5.408
Da'ira ja mai farin tsakiya da wani mashi mai launin rawaya ya ratsa shi.	1	male	7.2
Goma sha daya sai dari da goma= ya zama dubu daya da dari biyu da goma.	1	male	6.944
Wata korama ce a tsakiyar wasu duwatsu. Ga ciyawa ta tsaro a kan duwatsun, sannan ga hadari na alamun za a yi ruwan sama.	1	male	10.304
Goma sau goma= ya zama dari daya.	1	male	4.608
Taswirar wasu mutane guda uku manya biyu mace da namiji da suna tsaye da kuma ƙaramin yaro a tsakaninsu.	1	male	10.592
Daya a raba sau goma a raba sau miliyan daya = ya zama dubu dari kenan	1	male	5.504
Zane mai siffar gwangwani mai launin ruwan kasa da zane siffar zuciya mai launin ruwan toka a Samansa.	1	male	9.568
Ga wata gada da fitillun hanya guda biyu a gefe masu launin rawaya da ruwa yana wucewa ta Karkashinta.	1	male	8.64
A yau za a wuni ne gari wasai ba sanyi ba zafi, babu kuma ruwan sama.	1	male	6.208
Za a wuni cikin ni'mar ruwan sama babu hasken rana.	1	male	4.224
Za a kasance cikin yanayi na iska. Sai a sanya gilashi da takunkumin rufe baki da hanci.	1	male	6.816
Murabba'i mai launin algashi da zane mai kusurwowi hudu masu launin ruwan toka guda uku a Tsakiyarsa	1	male	10.176
Takwas sau Tara = ya zama Saba'in da biyu.	1	male	4.704
Bakwai sau takwas= ya zama hamsin da shida.	1	male	4.512
Lemar kare ruwan sama da rana mai launukan fari da ja da shudi da kuma kore. Sannan mai bakin mariki.	1	male	11.136
Sittin a raba wa mutum biyu= ya zama talatin.	1	male	5.408
A nan wani mutum ne sanye da riga mai launin bakar kasa da wando launin rawaya yana zaune a mai launin ruwan kasa yana karanta jarida.	1	male	12.448
Kofuna guda hudu dore a kan teburi mai launin ruwan kasa. Kofunan masu launukan rawaya da ja da ruwan toka da ruwan bakar kasa.	1	male	12.736
Za a samu hasken rana a wannan wunin gaba daya hade da iska-iska ba wacce take da karfi ba. Daga karfe bakwai na safiya har karfe biyar na yammaci.	1	male	10.464
A nan kuwa, za a tashi da hasken rana har zuwa karfe goma sha biyu. Haka kuma za a wuni.	1	male	7.2
Shudin da'ira hade da zane mai kusurwowi hudu mai tsini sama da Ƙasa (rhombus) guda biyu masu launin rawaya a gefe dama da hagu manne a jiki	1	male	12.576
Yau za a wuni ne ciki ni'mar gari, daga kashe kuma ana iya sami yayyafi ln ruwan sama.	1	male	7.232
Goma sha daya a haɗa da uku= ya zama goma sha hudu.	1	male	4.704
Hudu sau biyar = ya zama ashirin.	1	male	4.512
Wannan hasashen yanayi ne na mako guda tun daga Laraba zuwa wata. A sama dai ma'aunin yanayin digiri goma sha biyu ya nuna. Wato za a yi ta samun sanyi ne irin na raba da safe. Daga rana zuwa yamma a samu zafin rana.	1	male	20.704
Hasashen yanayi na dare da wurin zai yi tsit gari ya yi dum.	1	male	5.984

End of preview. Expand in Data Studio

Unified Hausa Speech Dataset v5

Dataset Description

A large-scale, cleaned, deduplicated, and quality-filtered Hausa speech dataset compiled from multiple open-source collections. Designed for Text-to-Speech (TTS) and Automatic Speech Recognition (ASR) research.

All audio is 16 kHz mono FLAC, silence-trimmed, loudness-normalized to -20 dBFS, and sorted by speaker_id so that all clips from the same speaker appear consecutively.

Dataset Summary

Metric	Train	Test	Total
Total Clips	776,724	7,419	784,143
Total Duration	697.02 hrs	7.86 hrs	704.88 hrs
Unique Speakers	2,058	31	2,089
Avg Clip Duration	3.231s	3.812s	—

Dataset Structure

Features

Column	Type	Description
`audio`	`Audio` (16 kHz)	Mono FLAC audio, silence-trimmed, loudness-normalized
`text`	`string`	NFC-normalized Hausa transcript
`speaker_id`	`int64`	Unique speaker identifier (small sequential integer, stable across runs)
`gender`	`string`	`"male"`, `"female"`, or `"other"`
`duration`	`float32`	Clip duration in seconds

Splits

Split	Description
`train`	WaxalNLP, NaijaVoices (3 batches), BibleTTS train, TWB Voice train, AfriVoices
`test`	BibleTTS test + dev (held-out), TWB Voice dev + test — never mixed into train

The dataset is sorted by speaker_id within each split, so all clips from the same speaker are grouped together. This is useful for speaker-adaptive TTS training and speaker embedding extraction.

Duration Analysis

Per-Split Statistics

Statistic	Train	Test
Min	1.000s	1.024s
Max	29.888s	29.792s
Mean	3.231s	3.812s
Median	2.656s	3.040s
Std	2.111s	2.678s

Duration Distribution (Train)

Range	Clips	Percentage
< 2 seconds	205,912	26.5%
2 — 5 seconds	471,012	60.6%
5 — 10 seconds	87,292	11.2%
10 — 20 seconds	11,551	1.5%
20 — 30 seconds	957	0.1%

Speaker Analysis

Top 20 Speakers by Sample Count (Train)

Rank	Speaker ID	Gender	Clips	Total Duration	Avg Duration
1	1888	male	39,300	5078.1 min	7.75s
2	139	female	8,364	422.3 min	3.03s
3	88	male	8,357	416.9 min	2.99s
4	217	female	8,353	417.1 min	3.00s
5	69	female	8,339	405.3 min	2.92s
6	225	male	8,339	515.1 min	3.71s
7	191	female	8,334	458.5 min	3.30s
8	189	female	8,327	425.7 min	3.07s
9	40	female	8,324	553.9 min	3.99s
10	86	male	8,319	390.6 min	2.82s
11	247	male	8,312	378.1 min	2.73s
12	79	female	8,286	416.5 min	3.02s
13	141	male	8,237	375.0 min	2.73s
14	159	female	8,232	302.8 min	2.21s
15	11	female	7,648	321.3 min	2.52s
16	275	female	7,607	321.1 min	2.53s
17	55	female	7,529	360.8 min	2.88s
18	76	female	7,504	365.4 min	2.92s
19	134	female	7,354	327.7 min	2.67s
20	13	male	7,203	306.0 min	2.55s

Gender Distribution

By Number of Speakers

Gender	Train Speakers	Test Speakers
Male	1111	18
Female	947	13
Other	0	0

By Number of Clips

Gender	Train Clips	Test Clips
Male	411,878	3,120
Female	364,846	4,299
Other	0	0

Text Analysis

Statistic	Train	Test
Min Length	2 chars	6 chars
Max Length	1139 chars	329 chars
Mean Length	45.5 chars	50.8 chars
Median Length	40 chars	42 chars

Quality Filtering Summary

During processing, clips were evaluated against quality criteria. The following clips were rejected:

Filter Reason	Rejected Clips
No Speaker ID	0
Too Short (< 1.0s)	9,063
Too Long (> 30.0s)	10,039
Excessive Clipping (> 1%)	0
Empty After Processing	0
Processing Errors	0
Total Rejected	19,102
Total Kept	801,711

Sources

Source	Repository	Clips Kept	Output Splits
WaxalNLP	`google/WaxalNLP`	1,588	train
NaijaVoices	`naijavoices/naijavoices-dataset`	700,329	train
BibleTTS	`vpetukhov/bible_tts_hausa`	40,076	train + test
TWB Voice	`CLEAR-Global/TWB-Voice-1.0`	29,268	train + test
AfriVoices	`suleiman2003/afri-temp-data5`	30,450	train

Audio Specifications

Parameter	Value
Sample Rate	16,000 Hz
Channels	Mono
Format	FLAC (16-bit PCM)
Loudness Target	-20 dBFS
Silence Trimming	25 dB threshold
Min Duration	1.0s
Max Duration	30.0s
Max Clipping	1%

Usage

from datasets import load_dataset

# Load the full dataset
ds = load_dataset("suleiman2003/W_hausa_v7")

# Load a specific split
train = load_dataset("suleiman2003/W_hausa_v7", split="train")
test = load_dataset("suleiman2003/W_hausa_v7", split="test")

# Access a sample
sample = train[0]
audio_array = sample["audio"]["array"]      # numpy array
sample_rate = sample["audio"]["sampling_rate"]  # 16000
transcript  = sample["text"]
speaker     = sample["speaker_id"]
gender      = sample["gender"]
duration    = sample["duration"]

# Stream (memory-efficient for large datasets)
ds_stream = load_dataset("suleiman2003/W_hausa_v7", split="train", streaming=True)
for sample in ds_stream:
    # process sample ...
    pass

License

Apache 2.0

Downloads last month: 603