6KSFx

Procedural audio, a method of generating sound from scratch using computational processes, offers a promising approach to creating sound effects, often referred to as "digital Foley". Despite its potential, research and optimisation of procedural audio has been limited by a lack of publicly available samples and models. This paper presents a dataset of synthetic audio samples designed to advance research in sound synthesis. By providing access to different synthesis methods and facilitating the development of evaluation techniques, this dataset aims to enable researchers, audio developers and sound designers to optimise and extend procedural audio applications. We are releasing 6,000 synthetic samples as a public dataset, a total of 2.56 GB and divided into 30 categories each one containing 100 samples . Each sample is named in the following format: "Name of the sound category" - "Sample number" - "Label" Sound Categories Label Sound Categories Label Sound Categories Label Sound Categories Label Applause 1 Gunshots 19 Whoosh 38 Twang 56 Church Bells 3 Helicopter 21 Fireworks 40 Bounce Rubber 58 Bubbles 5 PouringHotWater 23 Concrete Footsteps 42 Electric Buzz 60 Droplets 7 Rain 25 Wood Footsteps 44 Crackling 9 Thunder 27 Space Cannon 46 Debris Glass 11 Waves 29 Spray 48 Engine 13 Wind 31 Metal Debris 50 Explosions 15 Boat 33 Rocket 52 Fire 17 Jet 35 Concrete Debris 54 50 noise-added samples have already been included for machine learning applications.