Processed output from ARTPARK-IISc/Vaani: transcript-not-null filtered, reorganized by language. Hours computed from WAV duration in each parquet (run compute_vaani_hours.py to regenerate).

Non-null hours per language

Language	Hours
Hindi	963.01
Bengali	154.39
Telugu	151.49
Kannada	151.42
Marathi	68.46
Odia	52.48
Chakma	49.81
Garo	47.33
Malayalam	41.90
Nepali	38.17
Assamese	30.35
Tamil	24.34
Bhojpuri	23.78
Nagamese	23.76
English	22.84
Punjabi	21.71
Mizo	20.75
Maithili	19.79
Gujarati	19.48
Chhattisgarhi	17.61
Rajasthani	12.08
Wancho	11.91
Marwari	9.04
Garhwali	8.80
Magahi	6.00
Bajjika	5.06
Konkani	3.99
Kokborok	3.81
Angika	3.59
Khortha	3.22
Tulu	3.05
Kumaoni	2.55
Halbi	2.21
Urdu	1.88
Sumi	1.54
Khariboli	1.37
Idu_mishmi	1.08
Malvani	1.04
Sadri	0.96
Bundeli	0.82
Karbi	0.63
Surgujia	0.63
Rengma	0.51
Ao	0.49
Manipuri	0.45
Gondi	0.42
Magadhi	0.41
Chakhesang	0.38
Tangkhul	0.30
Rongmei	0.29
Awadhi	0.28
KhorthKhotta	0.26
Haryanvi	0.25
Surjapuri	0.25
Kurukh	0.23
Sambalpuri	0.21
Jaipuri	0.20
Khorth	0.20
Bearybashe	0.18
Bhili	0.16
Powari	0.13
Tenyidie	0.13
Sangtam	0.12
Santali	0.12
Tagin	0.12
Yimchunger	0.12
Nyishi	0.10
Oriya	0.08
Shekhawati	0.08
Kurmali	0.07
Bagri	0.06
Kashmiri	0.06
MagadhiMagahi	0.06
Sylheti	0.06
Thadou	0.06
Zeme	0.06
Bhatri	0.05
Galo	0.05
Hajong	0.05
Lotha	0.05
Angami	0.04
Liangmei	0.04
Mewari	0.04
Mewati	0.04
Sindhi	0.04
Wagdi	0.03
Harauti	0.02
Lambani	0.02
Phom	0.02
Rajbanshi	0.02
Vaiphei	0.02
Bihari	0.01
NissiDafla	0.01
Agariya	0.00
Kuki	0.00
Liangmai	0.00
Meitei	0.00

Total: 2,035.58 h (non-null transcript subset)

Dialect / spelling normalization (clubbing)

These names are variants or dialects that can be clubbed under one canonical language:

Variant (folder name)	Canonical language	Note
Oriya	Odia	Same language; Odia is the current official name.
Magadhi, MagadhiMagahi	Magahi	Magadhi/Magahi same language group.
Khorth, KhorthKhotta	Khortha	Spelling/dialect variants.
Idu_mishmi	Idu_Mishmi	Name fix (underscore → consistent casing).
NissiDafla	Nyishi	Nissi/Dafla are older names for Nyishi.
Liangmei	Liangmai	Spelling variant.
Bearybashe	Beary	Dialect name normalization.

Mapping is in vaani_language_map.json. To merge variant folders into the canonical language folders (and optionally remove variant folders), run:

/mnt/vdc/.venv/bin/python /mnt/vdc/merge_vaani_dialects.py --data-dir /mnt/vdc/vaani_data

Folder layout

<Language>/train-00000.parquet, train-00001.parquet, … — sharded parquet (max 500k rows per shard).
Each row has: audio, language, state, district, transcript, speakerID, gender, etc.

Full language list (pre-merge)

Agariya, Angami, Angika, Ao, Assamese, Awadhi, Bagri, Bajjika, Bearybashe, Bengali, Bhatri, Bhili, Bhojpuri, Bihari, Bundeli, Chakhesang, Chakma, Chhattisgarhi, English, Galo, Garhwali, Garo, Gondi, Gujarati, Hajong, Halbi, Harauti, Haryanvi, Hindi, Idu_mishmi, Jaipuri, Kannada, Karbi, Kashmiri, Khariboli, Khorth, KhorthKhotta, Khortha, Kokborok, Konkani, Kuki, Kumaoni, Kurmali, Kurukh, Lambani, Liangmai, Liangmei, Lotha, Magadhi, MagadhiMagahi, Magahi, Maithili, Malayalam, Malvani, Manipuri, Marathi, Marwari, Meitei, Mewari, Mewati, Mizo, Nagamese, Nepali, NissiDafla, Nyishi, Odia, Oriya, Phom, Powari, Punjabi, Rajasthani, Rajbanshi, Rengma, Rongmei, Sadri, Sambalpuri, Sangtam, Santali, Shekhawati, Sindhi, Sumi, Surgujia, Surjapuri, Sylheti, Tagin, Tamil, Tangkhul, Telugu, Tenyidie, Thadou, Tulu, Urdu, Vaiphei, Wagdi, Wancho, Yimchunger, Zeme.

Downloads last month: 20