| A | B | C | D | E | F | |
|---|---|---|---|---|---|---|
1 | ||||||
2 | Data Type | Languages | Description | Samples | ||
3 | Natural Conversational Voice Dataset | Hindi and Indian English | 10,885+ Hours of Hindi and 10,392+ of Indian English Conversation Voice Dataset. Natural two - person conversations recorded in each language. Each sample captures genuine interactions covering a range of topics and emotions typical in everyday conversations. With 1,000+ hours transcribed for each. | Hindi Audio | Hindi Sample Transcript | |
4 | Indian English Audio | English Sample Transcript | ||||
5 | Natural Conversational Voice Dataset | Telugu, Malayalam, Bengali, Punjabi, Tamil, Marathi | 10,000+ Hours of Telugu, Malayalam, Bengali, Punjabi, Tamil and Marathi. Natural two - person conversations recorded in each language. Each sample captures genuine interactions covering a range of topics and emotions typical in everyday conversations. Transcription not readily available. | Telugu Audio | Telugu Sample Transcript | |
6 | Malayalam Audio | Malayalam Sample Transcript | ||||
7 | Bengali Audio | Bengali Sample Transcript | ||||
8 | Punjabi Audio | Punjabi Sample Transcript | ||||
9 | Natural Conversational Voice Dataset - Low Resource Languages | Bodo, Maithili, Bhojpuri and Haryanvi | 10,000+ Hours for each language. Natural two-person conversations recorded in each language. Each sample captures genuine interactions covering a range of topics and emotions typical in everyday conversations. Transcription not readily available. | Bodo Audio | Bodo Sample Transcript | |
10 | Maithili Audio | Maithili Sample Transcript | ||||
11 | Bhojpuri Audio | Bhojpuri Sample Transcript | ||||
12 | Haryanvi Audio | Haryanvi Sample Transcript | ||||
13 | Read Speech Hindi | Hindi | This dataset comprises 1.2 million recordings totaling 10,374 hours of Hindi content, based on 124,000 unique paragraphs. With contributions from over 40,000 speakers across India, each 20-30 second recording captures linguistic diversity. (also available in English, Tamil, Telugu, Marathi, Bangla) | Audio | Transcript | |
14 | Natural conversations - 4 person | Hindi, Indian English | 550+ hours: Natural, four-person conversational audio recordings where participants were prompted to discuss various topics. The conversations include a mix of Hindi and Indian English. (also available in Tamil, Telugu, Marathi, Bangla) | Audio | Transcript | |
15 | High Mother Tounge Influence, High Emotion Conversations | Hindi | 1069 hours:Indian English. Speakers with medium and high degree of mother tongue influence were selected. Speakers were given an imaginary situation and asked to be overly emotional. (also available in English, Tamil, Telugu, Marathi, Bangla) | Audio | ||
16 | Real World Conversation Indian English (uncontrolled environment) | Indian English | 20,542 hours: Natural conversations between 2 people learning to speak fluently in English with 1,000 hours transcribed. The conversations took place in an uncontrolled environment, including adverse conditions and may have background noise as is the case in real world environments. Transcription not readily available. | Audio | ||
17 | Unique Environment High Emotion Utterances | Hindi | 1099 hours: Conversational data in Indian English and Hindi. Each recording has high degree of emotion. Users were given prompts of situations such as getting a promotion for a job and instructed to be highly emotional. Each utterance is a short conversation ranging from 30 sec to 3 min long. A total of 10 emotions are covered across all conversations. The users were instructed to be in certain specific environments like on a busy road and 8 other environments. (also available in English, Tamil, Telugu, Marathi, Bangla) | Audio (Environment: washroom with echo, Emotion: excited | ||
18 | Code Switching Conversation Hindi - English | Hindi- English | 1000+ hours: 1259 unique speakers have 2-person unscripted conversations. Each recording includes conversational audio where speakers fluidly switch between Hindi and English, capturing the natural bilingual speech patterns common in multilingual communities. (also available in other languages paired with English) | Audio | ||
19 | Josh Talks YouTube Videos Speech Dataset | Hindi | 1534 hours of studio recorded spontaneous speech covering 8 languages. This audio data is a part of the library of Josh Talks content that is shared on YouTube. (also available in English, Tamil, Telugu, Marathi, Bangla, Malayalam) | Audio | Transcript | |
20 | Voice Assistant Prompts with Phonetic Transcriptions | Hindi | This dataset comprises 100,000 utterances recorded by 200 unique speakers from 68 districts across India each speaking 500 phonetically rich sentences. (also available in English, Tamil, Telugu, Marathi, Bangla) | Audio | Phonetic Transcript | |