Note 6,380 hrs, 22 langs, CC-BY-4.0 — primary ASR training

Note 1,280 hrs — conversational/expressive speech

Note 6.66K rows, 22 langs — evaluation benchmark

Note 125K rows, 22 langs — noisy/code-switched eval

Note 64 hrs, 6 langs — far-field/noisy real-world speech

Note 79 hrs, bhojpuri — low-resource noisy speech

Note 2,280 hrs, 12 langs, CC-BY-4.0 — read speech

Note ~1,000 hrs — speaker variation TTS/ASR

Note 1,704 hrs, 22 langs — TTS-quality augmented speech

Note Speech translation variant of IndicVoices

Note Cleaned subset of IndicVoices

Note Dialect/accent variation speech data

Note Educational spoken tutorials in Indic languages

Note Prime Minister's speech data for Indic languages

Note Compilation of Indic speech resources

Note Crowdsourced read speech — 10 Indic languages, CC-0

Note IIIT-H LTRC collected Indic speech across multiple languages

Note 4,200 hrs English+Hindi large-scale multilingual speech

Note Multi-domain Hindi speech combining STT, TTS, and mixed sources

Note 360 hrs Telugu ASR corpus with de-duplicated transcripts

Note Telugu ASR speech dataset

Note IISc-MILE Tamil ASR corpus — large-scale transcribed Tamil speech

Note 1,000 hrs Tamil ASR corpus with de-duplicated transcripts

Note Hindi-English bilingual ASR for Parakeet fine-tuning

Note Auto-generated YouTube transcripts from 21 Hindi channels

Note Hindi ASR dataset in WebDataset format

Note Hindi ASR benchmark evaluating STT models

Note Clean Hindi dairy-domain ASR dataset

Note Hindi-Hinglish business-domain ASR

Note 5,000 hrs Bengali ASR data aggregated from public sources

Note 500 hrs Bengali ASR corpus with de-duplicated transcripts

Note Bengali long-form ASR dataset

Note Bengali-English code-switched ASR dataset

Note Unified Kannada ASR dataset v1.0 from multiple sources

Note IISc-MILE Kannada ASR corpus

Note 360 hrs Kannada ASR corpus with de-duplicated transcripts

Note Malayalam ASR dataset

Note Combined Malayalam ASR dataset

Note Multi-corpus Malayalam speech dataset aggregated from 5 public sources

Note Malayalam ASR corpus

Note Punjabi ASR datasets collection (687 downloads)

Note Punjabi ASR dataset filtered from Shrutilipi

Note Marathi ASR dataset (f)

Note Marathi ASR data

Note Marathi sentence-level ASR dataset

Note Gujarati ASR dataset at 16kHz

Note 100 hrs Gujarati ASR dataset

Note Aggregated Odia audio datasets for ASR research

Note Odia-English ASR dataset

Note Hindi-English parallel speech for code-switched ASR

Note Synthetic Indic speech data

Note Synthetic speech for Indic languages

Note Multilingual deepfake speech dataset for Indian languages (real+synthetic)