Skip to content
dhanu docs
Browse pages

Tasks

Sound effects

Generate short audio — a click, a chime, a whoosh, an ambient loop — from a description, and retry until the take is right.

Ask for a sound and the assistant generates a short audio clip into the task: a button click, a notification chime, a whoosh, an explosion, footsteps, or an ambient loop to sit under something.

Where it runs: in a task, when sound generation is enabled for your deployment.

What the assistant can do

It canWhat that looks like
Generate a sound effectA new clip from a description, attached to the task
Regenerate oneA fresh take of an existing sound, as a new version, optionally with a refined prompt

Clips run from half a second to thirty seconds. Every take is kept as a version, so you can compare three attempts at the same chime rather than losing the first one.

Getting the take you want

Describe the sound, not the object. “A short, dry mechanical click with a slight metallic ring” beats “a button sound”.

Say how long and how dense. “Two seconds”, “a seamless loop”, “sparse, with space between the drops”.

Retry rather than rephrase from scratch. Regeneration with a small adjustment — “same, but softer and shorter” — converges faster than a new description.

Sounds generated here are for prototypes, apps and demos. They are clips, not a scored soundtrack, and there is no music generation.