SpecSinGAN: Sound Effect Variation Synthesis Using Single-Image GANs
[Paper (Pre-print)]
Single-channel training data
Single-channel training data
Single-channel training data
Single-channel training data
All sounds Footsteps on concrete Footsteps on metal Gunshots Character jump Real 5.42 (1.33) 5.16 (1.32) 5.74 (1.32) 5.21 (1.33) 5.59 (1.28) Nemisindo 2.27 (1.65) 3.33 (1.95) 1.44 (0.89) 2.05 (1.34) Single-channel spectrograms 3.93 (1.56) 3.68 (1.45) 3.96 (1.60) 3.97 (1.71) 4.13 (1.45) Multi-channel spectrograms 4.38 (1.50) 3.95 (1.45) 4.29 (1.46) 4.87 (1.56) 4.44 (1.39)
Variation (1 to 7) All sounds Footsteps on concrete Footsteps on metal Gunshots Character jump Real 5.23 (1.62) 4.55 (1.69) 5.81 (1.35) 4.51 (1.66) 6.06 (1.04) Nemisindo 2.93 (1.82) 3.76 (1.83) 3.00 (1.76) 2.04 (1.40) Single-channel spectrograms 2.80 (1.63) 3.13 (1.66) 3.09 (1.67) 2.87 (1.46) 2.13 (1.55) Multi-channel spectrograms 3.55 (1.82) 3.73 (1.85) 3.81 (1.73) 4.16 (1.66) 2.53 (1.60)
Footsteps on concrete sounds by:
Freesound user "d.n.audio.uk", licensed under CC BY 3.0: https://freesound.org/people/d.n.audio.uk/sounds/437150/
Freesound user "Kinoton", licensed under CC0 1.0: https://freesound.org/people/Kinoton/sounds/546827/
Footsteps on metal sounds by:
Freesound user "Eelke", licensed under CC BY 3.0: https://freesound.org/people/Eelke/sounds/462598/
Freesound user "Eelke", licensed under CC BY 3.0: https://freesound.org/people/Eelke/sounds/462599/
Gunshot sounds by:
Freesound user "dwightsabeast", licensed under CC0 1.0: https://freesound.org/people/dwightsabeast/sounds/395572/
Freesound user "klangfabrik", licensed under CC0 1.0: https://freesound.org/people/klangfabrik/sounds/232750/
Freesound user "michorvath", licensed under CC0 1.0: https://freesound.org/people/michorvath/sounds/427596/
Character jump sounds by:
Freesound user "AbuNijmeh", licensed under CC0 1.0: https://freesound.org/people/AbuNijmeh/sounds/319104//
Freesound user "LamaMakesMusic", licensed under CC0 1.0: https://freesound.org/people/LamaMakesMusic/sounds/403657/
Freesound user "DCSFX", licensed under CC BY-NC 3.0: https://freesound.org/people/DCSFX/sounds/366126/
Sound examples
Footsteps on concrete
Single-channel unconditional synthesis (concatenated sounds) Multi-channel training data
Multi-channel unconditional synthesis (concatenated sounds) Footsteps on metal
Single-channel unconditional synthesis (concatenated sounds) Multi-channel training data
Multi-channel unconditional synthesis (concatenated sounds) Gunshots
Single-channel unconditional synthesis (concatenated sounds) Multi-channel training data
Multi-channel unconditional synthesis (concatenated sounds) Character jump
Single-channel unconditional synthesis (concatenated sounds) Multi-channel training data
Multi-channel unconditional synthesis (concatenated sounds) Listening study raw results
Plausibility (1 to 7) Training sounds attribution