id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_kxfutgt6zfhobmwolnrtzr7dle	Chris Kiefer	Sample-level sound synthesis with recurrent neural networks and conceptors	2019	24	.pdf	application/pdf	10430	1061	57	requirements preclude the use of these models for training with longer sound samples. Keywords Sound synthesis, Machine learning, Reservoir computing, Conceptors, Dynamical Sample-level sound synthesis with recurrent neural networks and conceptors. Jaeger's original work with ESNs included examples of models being trained to output questions through the application of conceptor models to a standard sound synthesis A new method of conceptor-based sound synthesis is demonstrated, named conceptular techniques that allow training and exploitation of the models for sounds synthesis. sound synthesis models to create new sonic variations of the original training material, Conceptular synthesis works by subdividing the audio training data into a set of subsequences, and learning an RNN and set of conceptors that can regenerate these subsequences, with the intention of resynthesising the audio sample by recombining the Figure 2 A comparison of the original kick drum sample and the output of the trained CCRNN model.	./cache/work_kxfutgt6zfhobmwolnrtzr7dle.pdf	./txt/work_kxfutgt6zfhobmwolnrtzr7dle.txt