Freesound's APIv2 lets you filter sounds by perceptual qualities like brightness, hardness and depth — not just text

The Freesound API exposes automatically-extracted audio descriptors that can be used as search filters. Beyond keyword search and duration filters, SOURCE lists brightness, hardness, and depth as perceptual filter dimensions. In SOURCE’s implementation these are AudioCommons descriptors sent to the API as range filters (the code builds a ac_<descriptor>:[min TO max] filter clause). Rather than continuous sliders, SOURCE buckets each descriptor into low/mid/high ranges (roughly 0–33 / 33–66 / 66–100), so a learner can ask Freesound for e.g. ‘bright’ or ‘hard’ sounds without auditioning any of them — the filter narrows the result set before download. This is distinct from conventional sample-library browsing, which relies on user-added tags: perceptual descriptors are computed from the audio itself, so they apply even to untagged sounds.

Examples

SOURCE’s code maps a ‘brightness’ filter set to ‘high’ into the API query ac_brightness:[66 TO 100], appended to the text/duration/license filters before searching. A query for ‘city’ with high brightness returns metallic, high-frequency urban recordings.

Assessment

Name three non-textual filter dimensions SOURCE exposes for Freesound searches. Explain what ‘brightness’ means as an audio descriptor and how it differs from a user-applied tag. How does SOURCE turn a ‘high brightness’ choice into an API filter?