Speech Recognition

VoxCeleb is a large-scale audio-visual speech dataset built from YouTube interview clips, widely used to train and benchmark deep speaker recognition models for speaker verification, speaker identification, and robust “in-the-wild” voice AI.

Casual Conversations is a large scale multimodal (video + audio) benchmark dataset built to evaluate and audit computer vision and speech models for accuracy across diverse ages, genders, apparent skin tones, and lighting conditions.

Large-scale audio-visual dataset comprising speech clips with no interfering background signals.

These expressions are produced at two levels of emotional intensities (regular and strong) except for the neutral emotion that only contains regular intensity.

Each video contains some number of procedure steps to fulfill a recipe. All the procedure segments are temporal localized in the video with starting time and ending time. The distributions of 1) video duration, 2) number of recipe steps per video, 3) recipe segment duration and 4) number of words per sentence are shown below.

This corpus contains 15 spontaneous dialogues and multi-participant conversations by deaf signers, 10 of which were recorded in authentic settings like a deaf club and a bar, 5 were recorded in the lab.

Data were collected from 100 subjects, yielding over thousand instances of synchronized data

The package including the videos and the metadata is available for non-commercial, academic research.

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Mandarin spoken in Shanghai, China

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Romanian spoken in Romania

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Polish spoken in Poland

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Panjabi spoken in Pakistan

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Mongolian spoken in Mongolia

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Mandarin (Traditional) spoken in Taiwan

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Mandarin (simplified) spoken in China

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Lao spoken in Laos

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Kannada spoken in India

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Spanish spoken in Spain

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, German spoken in Switzerland

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, French spoken in Algeria

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Farsi/Persian spoken in Iran

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, English spoken in UAE

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, English spoken in Philippines

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, English spoken in Hong Kong

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, English spoken in Australia

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Dutch spoken in Netherland

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Dutch spoken in Belgium

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Spanish spoken in Mexico

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Spanish spoken in Spain

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Catalan spoken in Catalonia, Spain.

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Sinhalese spoken in Sri Lanka

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Vietnamese spoken in Vietnam

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Tamil spoken in India

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Singaporean-English spoken in Singapore

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Punjabi spoken in India

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Malay spoken in Malaysia

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Bahasa spoken in Indonesia

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Thai spoken in Thailand

Unscripted conversation between two people. Approx. Audio Duration (Range) - 15-60 minutes, Gujarati spoken in India.

Various UK Accents

Dataset is fully transcribed and timestamped.

Dataset is fully transcribed and timestamped.

Voice Commands in the English Language

Dataset is fully transcribed and timestamped.

Dataset is fully transcribed and timestamped.

Dataset is fully transcribed and timestamped.

The data set includes 2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two speakers.

The data set includes 2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two speakers.

The data set includes 2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two speakers.

The data set includes 2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two speakers.

The data set includes 2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two speakers.

2 hours of time-stamped and transcribed unscripted speech data (i.e. natural conversation) between two Irish speakers

This data set contains recordings of the wake word "Alexa" in Canadian French (fr_CA) (e.g., "Alexa, raconte-moi une blague.").

Wake word "Alexa" in EU Spanish (es_ES) (e.g., "Eh, Alexa, cuéntame un chiste."). Each participant has recorded on average 70 utterances (minimum 50, maximum 75). This data set contains the voice command only.

This data set contains recordings of the wake word "Alexa" in Mexican Spanish (es_MX) used in voice commands (e.g., "Oye Alexa, cuéntame un chiste.").

US English voice commands including the wake word "Hey Siri" from 103 participants of age 19-68.

Korean voice commands including the wake word "Hi Bixby" from 52 participants in Seoul.

US English voice commands including the wake word "OK Google" from 103 participants of age 19-68.