Google Researcher published the Speech command Dataset! I'm publishing only 14 subcategories of voice data in a 1sec period.

I'm doing preprocessing and generating JSON files and uploading the dataset! feel free to use it

Enjoy! And develop your key spotting Application Efficiently

Multilingual Call Center Speech Recognition Dataset: 10,000 Hours

Dataset Summary

10,000 hours of real-world call center speech recordings in 7 languages with transcripts. Train speech recognition, sentiment analysis, and conversation AI models on authentic customer support audio. Covers support, sales, billing, finance, and pharma domains

Dataset Features

📊 Scale & Quality

• 10,000 hours of inbound & outbound calls
• Real-world field recordings - no synthetic audio
• With transcripts and concise summaries

🎙️ Audio Specifications

• Format: Single-channel (mono) telephone speech
• Sample rate: 8,000 Hz
• Non-synthetic source audio

🌍 Languages (7)

English, Russian, Polish, French, German, Spanish, Portuguese - Non-English calls include English translation - Additional languages available on request: Swedish, Dutch, Arabic, Japanese, etc.

🏢 Domains

Support, Billing/Account, Sales, Finance/Account Management, Pharma - Each call labeled by domain - Speaker roles annotated (Agent/Customer)

Full version of dataset is availible for commercial usage - leave a request on our website Axonlabs to purchase the dataset 💰

Purpose and Usage Scenarios

• Automatic Speech Recognition, punctuation restoration, and speaker diarization on telephone speech
• Intent detection, topic classification, and sentiment analysis from customer-service dialogs
• Post-call concise summaries for QA/quality monitoring and CRM automation
• Cross-lingual pipelines (original → English) and multilingual support models

This dataset is designed for building and improving speech recognition systems. It features transcribed customer service calls from real interactions across 160+ industries, including retail, banking, telecom, logistics, healthcare, and entertainment. Calls are natural, unscripted, and emotion-rich — making the data especially valuable for training models that must interpret speech under real-world conditions.

Each dataset entry includes:

• Full call transcription (agent + customer dialogue)
• Human-written call summary
• Overall sentiment label: positive, neutral, or negative
• Metadata: call duration, caller location (city, state, country), timestamp
• Optional: company name and industry tag

Use this dataset to:

• Train speech-to-text models on real customer language patterns. -Benchmark or evaluate speech recognition tools in support settings
• Improve voice interfaces, chatbots, and IVR systems.
• Model tone, frustration cues, and escalation behaviors
• Support LLM fine-tuning for tasks involving spoken input.s

This dataset provides your speech recognition models with exposure to genuine customer conversations, helping you build tools that can listen, understand, and act in line with how people actually speak.

The larger the volume you purchase, the lower the price will be.

The BengaliSpeechRecognitionDataset (BSRD) is a comprehensive dataset designed for the development and evaluation of Bengali speech recognition and text-to-speech systems. This dataset includes a collection of Bengali characters and their corresponding audio files, which are generated using speech synthesis models. It serves as an essential resource for researchers and developers working on automatic speech recognition (ASR) and text-to-speech (TTS) applications for the Bengali language. Key Features: • Bengali Characters: The dataset contains a wide range of Bengali characters, including consonants, vowels, and unique symbols used in the Bengali script. This includes standard characters such as 'ক', 'খ', 'গ', and many more. • Corresponding Speech Data: For each Bengali character, an MP3 audio file is provided, which contains the correct pronunciation of that character. This audio is generated by a Bengali text-to-speech model, ensuring clear and accurate pronunciation. • 1000 Audio Samples per Folder: Each character is associated with at least 1000 MP3 files. These multiple samples provide variations of the character's pronunciation, which is essential for training robust speech recognition systems. • Language and Phonetic Diversity: The dataset offers a phonetic diversity of Bengali sounds, covering different tones and pronunciations commonly found in spoken Bengali. This ensures that the dataset can be used for training models capable of recognizing diverse speech patterns. • Use Cases: o Automatic Speech Recognition (ASR): BSRD is ideal for training ASR systems, as it provides accurate audio samples linked to specific Bengali characters. o Text-to-Speech (TTS): Researchers can use this dataset to fine-tune TTS systems for generating natural Bengali speech from text. o Phonetic Analysis: The dataset can be used for phonetic analysis and developing models that study the linguistic features of Bengali pronunciation. • Applications: o Voice Assistants: The dataset can be used to build and train voice recognition systems and personal assistants that understand Bengali. o Speech-to-Text Systems: BSRD can aid in developing accurate transcription systems for Bengali audio content. o Language Learning Tools: The dataset can help in creating educational tools aimed at teaching Bengali pronunciation.

…………………………………..Note for Researchers Using the dataset………………………………………………………………………

This dataset was created by Shuvo Kumar Basak. If you use this dataset for your research or academic purposes, please ensure to cite this dataset appropriately. If you have published your research using this dataset, please share a link to your paper. Good Luck.

The LibriSpeech dataset is a corpus of read English speech. It was created by Google and Carnegie Mellon University, and is used for training and evaluating speech recognition models. The dataset consists of 960 hours of speech, divided into a training set of 920 hours and a test set of 40 hours. The speech is read by a variety of speakers, and covers a wide range of topics. The dataset is available for free download.

Introduction

Introducing the UK English Scripted Monologue Speech Dataset for the Healthcare Domain, a voice dataset built to accelerate the development and deployment of English language automatic speech recognition (ASR) systems, with a sharp focus on real-world healthcare interactions.

Speech Data

This dataset includes over 6,000 high-quality scripted audio prompts recorded in UK English, representing typical voice interactions found in the healthcare industry. The data is tailored for use in voice technology systems that power virtual assistants, patient-facing AI tools, and intelligent customer service platforms.

Topic Coverage

The prompts span a broad range of healthcare-specific interactions, such as:

Contextual Depth

To maximize authenticity, the prompts integrate linguistic elements and healthcare-specific terms such as:

These elements make the dataset exceptionally suited for training AI systems to understand and respond to natural healthcare-related speech patterns.

Transcription

Every audio recording is accompanied by a verbatim, manually verified transcription.

The Aurora project was originally set up to establish a worldwide standard for the feature extraction software which forms the core of the front-end of a DSR (Distributed Speech Recognition) system. The AURORA-5 database has been mainly developed to investigate the influence on the performance of automatic speech recognition for a hands-free speech input in noisy room environments. Furthermore two test conditions are included to study the influence of transmitting the speech in a mobile communication system.The earlier three Aurora experiments had a focus on additive noise and the influence of some telephone frequency characteristics. Aurora-5 tries to cover all effects as they occur in realistic application scenarios. The focus was put on two scenarios. The first one is the hands-free speech input in the noisy car environment with the intention of controlling either devices in the car itself or retrieving information from a remote speech server over the telephone. The second one covers the hands-free speech input in a type of office or in a type of living room to control e.g. a telephone device or some audio/video equipment.The AURORA-5 database contains the following data:•Artificially distorted versions of the recordings from adult speakers in the TI-Digits speech database downsampled at a sampling frequency of 8000 Hz. The distortions consist of: - additive background noise, - the simulation of a hands-free speech input in rooms, - the simulation of transmitting speech over cellular telephone networks.•A subset of recordings from the meeting recorder project at the International Computer Science Institute. The recordings contain sequences of digits uttered by different speakers in hands-free mode in a meeting room.•A set of scripts for running recognition experiments on the above mentioned speech data. The experiments are based on the usage of the freely available software package HTK where HTK is not part of this resource.Further information is also available at the following address: http://aurora.hsnr.de

An audio dataset of spoken words designed to help train and evaluate keyword spotting systems. Its primary goal is to provide a way to build and test small models that detect when a single word is spoken, from a set of ten target words, with as few false positives as possible from background noise or unrelated speech. Note that in the train and validation set, the label "unknown" is much more prevalent than the labels of the target words or background noise. One difference from the release version is the handling of silent segments. While in the test set the silence segments are regular 1 second files, in the training they are provided as long segments under "background_noise" folder. Here we split these background noise into 1 second clips, and also keep one of the files for the validation set.

To use this dataset:

import tensorflow_datasets as tfds

ds = tfds.load('speech_commands', split='train')
for ex in ds.take(4):
 print(ex)

See the guide for more informations on tensorflow_datasets.

The M-AILABS Speech Dataset is the first large dataset that we are providing free-of-charge, freely usable as training data for speech recognition and speech synthesis.

Most of the data is based on LibriVox and Project Gutenberg. The training data consist of nearly thousand hours of audio and the text-files in prepared format.

A transcription is provided for each clip. Clips vary in length from 1 to 20 seconds and have a total length of approximately shown in the list (and in the respective info.txt-files) below.

The texts were published between 1884 and 1964, and are in the public domain. The audio was recorded by the LibriVox project and is also in the public domain – except for Ukrainian.

Ukrainian audio was kindly provided either by Nash Format or Gwara Media for machine learning purposes only (please check the data info.txt files for details).

== Quick facts ==

The most up-to-date and comprehensive podcast database available All languages & All countries Includes over 3,600,000 podcasts Features 35+ data fields , such as basic metadata, global rank, RSS feed (with audio URLs), Spotify links, and more Delivered in SQLite format Learn how we build a high quality podcast database: https://www.listennotes.help/article/105-high-quality-podcast-database-from-listen-notes

== Use Cases ==

AI training, including speech recognition, generative AI, voice cloning / synthesis, and news analysis Alternative data for investment research, such as sentiment analysis of executive interviews, market research and tracking investment themes PR and marketing, including social monitoring, content research, outreach, and guest booking ...

== Data Attributes ==

See the full list of data attributes on this page: https://www.listennotes.com/podcast-datasets/fields/?filter=podcast_only

How to access podcast audio files: Our dataset includes RSS feed URLs for all podcasts. You can retrieve audio for over 170 million episodes directly from these feeds. With access to the raw audio, you’ll have high-quality podcast speech data ideal for AI training and related applications.

== Custom Offers ==

We can provide custom datasets based on your needs, such as language-specific data, daily/weekly/monthly update frequency, or one-time purchases.

We also provide a RESTful API at PodcastAPI.com

Contact us: hello@listennotes.com

== Need Help? ==

If you have any questions about our products, feel free to reach out hello@listennotes.com

== About Listen Notes, Inc. ==

Since 2017, Listen Notes, Inc. has provided the leading podcast search engine and podcast database.

========================================================

This dataset is developed by Md Ashraful Islam, SUST CSE'2010 Md Mahadi Hasan Nahid, SUST CSE'2010 (nahid-cse@sust.edu)

Department of Computer Science and Engineering (CSE) 
Shahjalal University of Science and Technology (SUST), www.sust.edu 

Special Thanks To Mohammad Al-Amin, SUST CSE'2011 Md Mazharul Islam Midhat, SUST CSE'2010 Md Mahedi Hasan Nayem, SUST CSE'2010
Avro Key Board, Omicron lab, https://www.omicronlab.com/index.html

=========================================================

It is a Audio Text Parallel Corpus. This dataset contains Some Recording Audio of Bangla Real Number and Its Coresponding Text. Specially designed for Bangla Speech recognition.

There are five speakers(alamin, ashraful, midhat, nahid, nayem) in this dataset.

Vocabulary Contains only bangla real numbers (shunno-ekshoto, hazar, loksho, koti, doshomic etc.)

Total Number of Audio file : 175 (35 from each speaker) Age range of the speakers : 20-23

Total Size: 32.4MB

TextData.txt file contains the text of the audio set. Each line starts with tag and ends with tag. The file name is added after each line using parenthesis, in this audio file you will get its recorder Audio Data. This text data actually generated using Avro (Free Opensourse Writting Software).

==========================================================

For Full Data: please contact nahid-cse@sust.edu

This dataset contains common speech and noise corpora for evaluating fundamental frequency estimation algorithms as convenient JBOF dataframes. Each corpus is available freely on its own, and allows redistribution:

CMU-ARCTIC (BSD license) [1]

FDA (free to download) [2]

KEELE (free for noncommercial use) [3]

MOCHA-TIMIT (free for noncommercial use) [4]

PTDB-TUG (ODBL license) [5]

NOISEX (free to download) [7]

QUT-NOISE (CC-BY-SA license) [8]

These files are published as part of my dissertation, "Pitch of Voiced Speech in the Short-Time Fourier Transform: Algorithms, Ground Truths, and Evaluation Methods", and in support of the Replication Dataset for Fundamental Frequency Estimation.

References:

John Kominek and Alan W Black. CMU ARCTIC database for speech synthesis, 2003.

Paul C Bagshaw, Steven Hiller, and Mervyn A Jack. Enhanced Pitch Tracking and the Processing of F0 Contours for Computer Aided Intonation Teaching. In EUROSPEECH, 1993.

F Plante, Georg F Meyer, and William A Ainsworth. A Pitch Extraction Reference Database. In Fourth European Conference on Speech Communication and Technology, pages 837–840, Madrid, Spain, 1995.

Alan Wrench. MOCHA MultiCHannel Articulatory database: English, November 1999.

Gregor Pirker, Michael Wohlmayr, Stefan Petrik, and Franz Pernkopf. A Pitch Tracking Corpus with Evaluation on Multipitch Tracking Scenario. page 4, 2011.

John S. Garofolo, Lori F. Lamel, William M. Fisher, Jonathan G. Fiscus, David S. Pallett, Nancy L. Dahlgren, and Victor Zue. TIMIT Acoustic-Phonetic Continuous Speech Corpus, 1993.

Andrew Varga and Herman J.M. Steeneken. Assessment for automatic speech recognition: II. NOISEX-92: A database and an experiment to study the effect of additive noise on speech recog- nition systems. Speech Communication, 12(3):247–251, July 1993.

David B. Dean, Sridha Sridharan, Robert J. Vogt, and Michael W. Mason. The QUT-NOISE-TIMIT corpus for the evaluation of voice activity detection algorithms. Proceedings of Interspeech 2010, 2010.

The Accented English Speech Dataset provides over 1,000 hours of authentic conversational recordings designed to strengthen ASR systems, conversational AI, and voice applications. Unlike synthetic or scripted datasets, this collection captures real human-to-human and chatbot-guided dialogues, reflecting natural speech flow, spontaneous phrasing, and diverse accents.

Off-the-shelf recordings are available from:

Mexico Colombia Guatemala Costa Rica El Salvador Dominican Republic South Africa

This ensures exposure to Latin American, Caribbean, and African English accents, which are often missing from mainstream corpora.

Beyond these, we support custom collection in any language and any accent worldwide, tailored to specific project requirements.

Audio Specifications

Format: WAV Sample rate: 48kHz Sample size: 16-bit PCM Channel: Mono/Stereo Double-track recording: Available upon request (clean separation of speakers) Data Structure and Metadata Dual-track or single-channel audio depending on project need Metadata includes speaker ID, demographic attributes, accent/region, and context Dialogues include both structured (chatbot/task-based) and free-flow natural conversations

Use Cases

• ASR Training & Benchmarking – Improve transcription across accented English
• Accent Adaptation – Build robust, inclusive systems that work in real-world scenarios
• Multilingual Voice Interfaces – Expand IVR and assistants to support more voices
• Conversational AI – Train chatbots on authentic, unstructured dialogue
• Voice Biometrics – Support research in identity verification and speaker profiling
• Model Fine-Tuning – Enrich foundation models with high-quality speech data

Why It Matters

Mainstream datasets disproportionately focus on U.S. and U.K. English. This dataset fills the gap with diverse accented English coverage, and the ability to collect any language or accent on demand, enabling the creation of fairer, more accurate, and globally deployable AI solutions.

Key Highlights

• 1,000+ hours of accented English speech
• Ready-to-use coverage from Latin America, Caribbean, and Africa
• Authentic dialogues: human-to-human and chatbot-guided
• WAV, 48kHz, 16-bit PCM, mono/stereo, double-track option
• Metadata-rich recordings for advanced AI research
• Custom collection in any language and accent

This dataset consists of audio recordings in Indonesian language, categorized into two distinct classes: human voices (real) and synthetic voices generated using artificial intelligence (AI). Each class comprises 21 audio files, resulting in a total of 42 audio files. Each recording has a duration ranging from approximately 4 to 9 minutes, with an average length of around 6 minutes per file. All recordings are provided in WAV format and accompanied by a CSV file containing detailed duration metadata for each audio file.

This dataset is suitable for research and applications in speech recognition, voice authenticity detection, audio analysis, and related fields. It enables comparative analysis between natural Indonesian speech and AI-generated synthetic speech.

This contain Egyptian Arabic Dataset for MTC Competition, be careful version 4 for training and validation and 5 are used for testing. The code for the ASR can be found on our notebooks, for any advice or material feel free to contact us. Best of luck guys!!

Introduction

The Tamil Scripted Monologue Speech Dataset for the General Domain is a carefully curated resource designed to support the development of Tamil language speech recognition systems. This dataset focuses on general-purpose conversational topics and is ideal for a wide range of AI applications requiring natural, domain-agnostic Tamil speech data.

Speech Data

This dataset features over 6,000 high-quality scripted monologue recordings in Tamil. The prompts span diverse real-life topics commonly encountered in general conversations and are intended to help train robust and accurate speech-enabled technologies.

Topic Coverage

The dataset covers a wide variety of general conversation scenarios, including:

Contextual Features

To enhance authenticity, the prompts include:

Each prompt is designed to reflect everyday use cases, making it suitable for developing generalized NLP and ASR solutions.

Transcription

Every audio file in the dataset is accompanied by a verbatim text transcription, ensuring accurate training and evaluation of speech models.

Metadata

Rich metadata is included for detailed filtering and analysis:

Applications & Use Cases

This dataset can power a variety of Tamil language AI technologies, including:

Arabic Speech Commands Dataset

This dataset is designed to help train simple machine learning models that serve educational and research purposes in the speech recognition domain, mainly for keyword spotting tasks.

Dataset Description

Our dataset is a list of pairs (x, y), where x is the input speech signal, and y is the corresponding keyword. The final dataset consists of 12000 such pairs, comprising 40 keywords. Each audio file is one-second in length sampled at 16 kHz. We have 30 participants, each of them recorded 10 utterances for each keyword. Therefore, we have 300 audio files for each keyword in total (30 * 10 * 40 = 12000), and the total size of all the recorded keywords is ~384 MB. The dataset also contains several background noise recordings we obtained from various natural sources of noise. We saved these audio files in a separate folder with the name background_noise and a total size of ~49 MB.

Dataset Structure

There are 40 folders, each of which represents one keyword and contains 300 files. The first eight digits of each file name identify the contributor, while the last two digits identify the round number. For example, the file path rotate/00000021_NO_06.wav indicates that the contributor with the ID 00000021 pronounced the keyword rotate for the 6th time.

Data Split

We recommend using the provided CSV files in your experiments. We kept 60% of the dataset for training, 20% for validation, and the remaining 20% for testing. In our split method, we guarantee that all recordings of a certain contributor are within the same subset.

License

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. For more details, see the LICENSE file in this folder.

Citations

If you want to use the Arabic Speech Commands dataset in your work, please cite it as:

@article{arabicspeechcommandsv1, author = {Ghandoura, Abdulkader and Hjabo, Farouk and Al Dakkak, Oumayma}, title = {Building and Benchmarking an Arabic Speech Commands Dataset for Small-Footprint Keyword Spotting}, journal = {Engineering Applications of Artificial Intelligence}, year = {2021}, publisher={Elsevier} }

This dataset is collected to enhance research into speech recognition systems for dysarthic speech.

This database is orignally created for a resource for developing advanced models in automatic speech recognition that are more suited to the needs of people with dysarthria.

OpenSLR is a site devoted to hosting speech and language resources, such as training corpora for speech recognition, and software related to speech recognition. They intend to be a convenient place for anyone to put resources that they have created, so that they can be downloaded publicly. They aim to provide a central, hassle-free place for others to put their speech resources. see there http://www.openslr.org/contributions.html

Supported Task

Automatic Speech Recognition

Languages… See the full description on the dataset page: https://huggingface.co/datasets/Pratik/Gujarati_OpenSLR.