Loading datasets...
License:
NOODL-1.0
Steward:
Institute of African Digital HumanitiesDataset ID:
cmrbacjej00lfmm07miqrz23s
Release Date: 7/7/2026
Format: MP3, TSV
Size: 169.22 MB
Share
Tupuri-Bango_TTS-Dataset (female voice) is a scripted speech dataset dedicated to the documentation and technological development of Tupuri (ISO 639-3: tui), a Chadic language of the Afro-Asiatic phylum spoken in the Kaele and Mayo-Danay Divisions of the Far-North Region of Cameroon and in adjacent areas of southern Chad. The dataset was compiled in the framework of the Mozilla Data Collective initiative (2026) and represents the Bango dialectal variety of Tupuri. Its principal added value lies in speaker gender: all 2,034 recordings are produced by a single female native speaker, in direct contrast to the previously released Sample Tupuri-Bango_TTS-Dataset (male voice) (https://mozilladatacollective.com/datasets/cmqf6bpns06gal20726pxld9z), which comprises exclusively male-voice recordings. Prior to this release, no female-voice speech resource existed for Tupuri-Bango. This dataset closes that gap, enabling the development of gender-balanced and multi-speaker TTS systems for the language, and making possible, for the first time, a direct comparison of male and female voice characteristics (pitch, prosody, tonal realisation) within the same dialectal variety and orthographic framework. The dataset comprises 2,034 high-quality MP3 audio recordings of Tupuri-Bango sentences read by the speaker across 21 recording sessions, together with per-session sentence-to-audio mapping files enabling precise alignment between textual and acoustic data. Sentences were drawn from a scripted speech prompt list and read in a controlled environment. The transcription of all sentences follows the General Alphabet of Cameroon's Languages (AGLC; French acronym: Alphabet Général des Langues Camerounaises), the reference standard for Cameroonian national languages. The Tupuri-Bango orthography employed in this dataset is characterised by an extended vowel inventory — including the open-mid front unrounded vowel ɛ and the open-mid back rounded vowel ɔ — a set of nasalized vowels represented by the tilde diacritic (ã, ẽ, ũ, õ), with long vowels encoded by vowel doubling (aa, ɛɛ, ɔɔ, ãã, etc.), a series of implosive consonants written with hooked letters (ɓ for the bilabial implosive and ɗ for the alveolar implosive), the velar nasal consonant ŋ (eng), a multi-register tone-marking system combining level (acute, grave) and contour (caron) diacritics applied to vowels, and the apostrophe (' or ') as a glottal stop or glottalization marker. The parallel availability of AGLC-transcribed text and aligned speech makes the dataset suitable for a wide range of applications, including text-to-speech (TTS) synthesis, automatic speech recognition (ASR), forced alignment, pronunciation modelling, and language learning tools. It also directly supports efforts to standardise and normalise the digital representation of Tupuri in language technology contexts.
Licensing
Nwulite Obodo Open Data Licence 1.0 (NOODL-1.0)
https://licensingafricandatasets.com/nwulite-obodo-licenseRestrictions/Special Constraints
By downloading this dataset, you agree: - To use it for research and scientific use only - That you will not re-host or re-share this dataset
Forbidden Usage
You agree not to use the data for: determining the identity of any speaker in the dataset; attempting to clone any voice or train models that imitate any speaker in this dataset; Generative AI; reproduction; duplication; modification; augmentation; copying; distribution; transmission; display; sale; transfer; publication or creation of derivative works without the explicit permission of the legal owner of the dataset.
Intended Use
(a) Speech-related tasks: - Text-to-speech (TTS) synthesis: The dataset provides clean sentence–audio pairs from a single female speaker across 21 recording sessions and is directly suited for training, fine-tuning, and evaluating speech synthesis models for Tupuri-Bango. Combined with the existing male-voice dataset, it enables multi-speaker and gender-balanced TTS model development, voice selection, and speaker-conditioned synthesis for Tupuri-Bango — capabilities that were not previously possible with a male-only speech resource. - Multi-speaker and cross-gender TTS research: As the first female-voice counterpart to the Sample Tupuri-Bango_TTS-Dataset (male voice), this dataset supports comparative studies of male/female voice characteristics (pitch range, prosody, tonal realisation) within the same dialect and orthographic system, and enables speaker embedding, voice-conversion, and speaker-adaptation research for an under-resourced tonal language. - Automatic speech recognition (ASR): Audio–text alignment enables the training and evaluation of speaker-diverse speech recognition models for Tupuri. Combined with the Tupuri-ASR-Dataset and Common Voice Scripted Speech 26.0 - Tupuri, this dataset contributes female-voice acoustic diversity that helps reduce gender bias in ASR performance for the language. - Speech–text alignment / forced alignment benchmarking: Fine-grained audio–text pairing provides ground truth for evaluating phoneme- or word-level aligners adapted to Chadic languages of the Far-North Region of Cameroon, across different speaker genders. - Pronunciation modelling: The AGLC-transcribed sentences, combined with aligned audio, provide a resource for developing grapheme-to-phoneme (G2P) models and pronunciation lexicons for Tupuri-Bango, informed by both male and female speech realisations when combined with the companion dataset. (b) Linguistic and lexicographic tasks: - Phonological analysis: The dataset enables systematic study of the phonological and tonal system of Tupuri-Bango, including its tonal contrasts, nasalized vowel inventory (ã, ẽ, ũ, õ), extended vowel inventory (ɛ, ɔ), implosive consonant series (ɓ, ɗ), and the behaviour of long vowels, as realised by a female speaker. - Dialectological research: As a dataset explicitly representing the Bango variety of Tupuri, the dataset provides a point of comparison for future documentation of other Tupuri varieties, including Tupuri-Banwere. - Orthographic standardisation and normalisation: The dataset can serve as a reference corpus for evaluating and training text normalisation models aligned with the AGLC standard for Tupuri. - Language documentation: The dataset contributes to the digital documentation of Tupuri-Bango scripted speech in AGLC orthography, extending prior male-voice documentation with a female-voice complement and supporting efforts to broaden the digital presence of this Chadic language of the Far-North Region of Cameroon.
Tupuri (also referred to as Toupouri or Tupure) is a Chadic language belonging to the Afro-Asiatic phylum. It is spoken primarily in the Kaele and Mayo-Danay Divisions of the Far-North Region of Cameroon and in adjacent areas of southern Chad. Despite its regional significance, Tupuri remains substantially underrepresented in language technology resources.
The dialectal situation of Tupuri is likely not fully documented. Two of its most salient varieties are:
Tupuri-Bango (represented by the present dataset)
Tupuri-Banwere
The writing system used for the transcription of Tupuri-Bango in this dataset is the General Alphabet of Cameroon's Languages (AGLC). The AGLC provides a phonologically motivated orthographic standard for Cameroonian national languages and serves as the reference framework for Tupuri literacy materials.
The vowel system attested in the dataset includes the following oral vowels:
a, e, i, o, u, ɛ, ɔ
Where:
ɛ (epsilon): open-mid front unrounded vowel
ɔ (open-o): open-mid back rounded vowel
Nasalized vowels are represented by the tilde diacritic placed directly on the vowel:
ã, ẽ, ũ, õ
Long vowels are represented by vowel doubling (e.g., aa, oo, ɛɛ, ɔɔ, ãã).
The consonant inventory reflected in the dataset includes simple and digraph consonants:
b, c, d, f, g, h, j, k, l, m, n, p, r, s, t, w, y
Implosive consonants (AGLC hooked letters):
ɓ: bilabial implosive (b with hook)
ɗ: alveolar implosive (d with hook)
Special symbols:
ŋ (eng): velar nasal consonant
' or ' (apostrophe): glottal stop or glottalization marker
Tupuri-Bango is a tonal language. The dataset employs systematic tone marking on vowels in accordance with the AGLC convention. The following diacritics are attested in the dataset:
Level tones:
High tone (H): acute accent — á, é, í, ú
Low tone (L): grave accent — à, è, ì, ù
Contour tones:
Rising tone (LH): caron — ě, ǐ, ǔ
Vowels without a tone diacritic are contextually interpreted in accordance with the AGLC convention for Tupuri.
The dataset was compiled from scripted speech prompt lists read by a native speaker. Sentences were selected to provide broad phonological coverage of Tupuri-Bango and were transcribed in accordance with the AGLC orthographic standard.
The dataset represents scripted speech in Tupuri-Bango, covering a broad range of everyday sentence types drawn from a general-purpose TTS/ASR prompt list. All utterances are scripted rather than spontaneous.
All 2,034 recordings in this dataset were produced by a single female native speaker of Tupuri-Bango. This is the distinguishing feature of the dataset relative to the existing Sample Tupuri-Bango_TTS-Dataset (male voice), which comprises exclusively male-voice recordings from a different speaker. No sentence content is shared by design between the two datasets' recording sessions beyond overlap inherent in the common prompt list.
Total audio duration: 13,745 seconds (03h 49m 05s), distributed across 2,034 MP3 audio clips in 21 recording sessions.
The dataset is organised into 21 recording sessions:
Session 01: 100 clips (9m 19s)
Session 02: 100 clips (9m 38s)
Session 03: 100 clips (12m 13s)
Session 04: 100 clips (10m 22s)
Session 05: 100 clips (12m 09s)
Session 06: 100 clips (14m 47s)
Session 07: 100 clips (11m 31s)
Session 08: 100 clips (11m 58s)
Session 09: 100 clips (11m 00s)
Session 10: 100 clips (11m 21s)
Session 11: 100 clips (11m 31s)
Session 12: 100 clips (10m 40s)
Session 13: 100 clips (12m 15s)
Session 14: 100 clips (8m 32s)
Session 15: 100 clips (11m 40s)
Session 16: 100 clips (10m 35s)
Session 17: 100 clips (9m 21s)
Session 18: 100 clips (10m 02s)
Session 19: 100 clips (12m 11s)
Session 20: 100 clips (13m 10s)
Session 21: 34 clips (4m 41s)
Each session folder contains:
MP3 audio clips
One per-session sentence-to-audio mapping file (mapping.tsv), with 4 columns
audio_filename: filename of the audio clip (MP3)
key: unique hash identifier of the recording
sentence: sentence text as read by the speaker, transcribed in AGLC orthography
attempts: number of recording attempts before acceptance
| audio file | sentence (Tupuri-Bango, AGLC) |
|---|---|
| afd84d3810cab1328c32615c733a9ba1.mp3 | Wɛr dɛfge ne masi pur sarsar ɓi wern wo |
| 4f9263c69b475bcad33f3f16beb7704f.mp3 | Ndi ko we, ndi mbi de kaage ma age de mo po no ga ma kay nday de deŋ ma ti' riŋ ? |
| d1f2554e575cfe591efeeea6efa2d810.mp3 | Carge ma po ɓi sir la kaŋ jar jɔŋ ne jar ma da'ge wɔsɛla la ti se ɓɛ. |
| b732348b88471dcb3ae2223df0ec3fff.mp3 | Podge targe ma kani fu̧ŋ jag wɔ ha̧a̧ ti naw 28 ɓil Fɛw Darge. |
| 546bc1dcadc75fdd4b1634e412d8d80f.mp3 | De maaga tuu tee wɛ gɔ no,laa cuu ɓil ɓɛ,blam ɓɛ kaŋ koo ciŋ ti ɓɛ. |