Click on the blue links below to preview 5 seconds extracted from the selected sequence. The audio track used in each preview is the first channel of the microphone array of AVA1

conversation1_t1

conversation1_t2

conversation1_t3

conversation1_t4

femalemonologue1_t1

femalemonologue1_t2

femalemonologue1_t3

femalemonologue1_t4

femalemonologue2_t1

femalemonologue2_t2

femalemonologue2_t3

femalemonologue2_t4

interactive1_t1

interactive1_t2

interactive1_t3

interactive2_t1

interactive2_t2

interactive2_t3

interactive2_t4

interactive3_t1

interactive4_t1

interactive4_t2

interactive4_t3

malemonologue1_t1

malemonologue1_t2

malemonologue1_t3

malemonologue1_t4

malemonologue2_t1

malemonologue2_t2

malemonologue2_t3


If you have received the username and password, you can download the TragicTalkers dataset.


Acknowledgments

This work is supported by InnovateUK (105168) ‘Polymersive: Immersive video production tools for studio and live events’ and a PhD studentship from the Doctoral College of the University of Surrey. Thanks to actors Phoebe Salem and Mason Stickland for their time and availability; to Mohd Azri Mohd Izhar, Hansung Kim, and Charles Malleson for their contribution to the audio-visual recordings; to Umar Marikkar for developing utility scripts supporting easy access to the data, e.g., data loaders and audio-visual feature extractors; to Alexander Todd for helping during the laser cut of the AVA Rigs.