Sound Innovation Engines: Synthetic Audio from Quality Diversity Search

Björn Þór Jónsson published this collection in 2024 to support research on generative sound synthesis using Quality Diversity algorithms. The data includes synthetic sound objects and rendered audio files evolved via Pattern-Producing Networks and guided by the YAMNet classifier.

Use Cases

• Evaluating the diversity of sounds generated by Pattern-Producing Networks
• Analyzing the accuracy of YAMNet classifier labels on synthetic audio
• Benchmarking Quality Diversity algorithms in the context of sonic design

Strengths

• Includes rendered sound files for immediate auditory analysis
• Utilizes YAMNet-guided labels for sound classification
• Provides outputs from a specific Quality Diversity search algorithm

Limitations

• Unknown record count and file size
• Synthetic nature may not represent acoustic instrument nuances
• Lack of column metadata for structured analysis

Provenance

Source

DataverseNO

Collection Method

synthetic generation

Freshness

Last updated March 2024