Adjust detected sound groups inside one recording

How the tool works

When the separation model is available, the tools offer original and processed listening with an explicit Apply or Confirm step. Supported sound groups share mixer controls; a detailed detection does not imply its own independent slider. Processing quality depends on the recording, and the original remains available.

Build it step by step

Open Audio for the source and choose Identify sounds. If the separation model is unavailable, review the detected list and original playback without expecting category controls to work.

• Compare the detected list with the original and identify which sound group needs a change.
• Wait for supported groups to prepare; detailed labels may share one control instead of individual sliders.
• Change one group at a time and leave other levels stable so you can hear its effect.
• Use Listen to changes to render the draft and compare speech, background, and the residual sound.
• Confirm only a useful mix, or cancel to discard the draft without changing the project.

An example to adapt

Illustrative exercise: a short recording contains speech and a dog sound. Lower the supported dog group modestly, render a draft, and check that quiet speech still sounds complete.

Listen to Other sounds as part of the whole result, not as a list of reliably separated leftovers. If the voice becomes thin or the bark remains distracting, keep the original instead of stacking further changes without comparison.

Check before finishing

The full original source is currently limited to 60 seconds. Discovery can analyze a longer selected range, but that does not extend mixer preparation. Some detections have no extraction controls, and overlapping estimates can leave artifacts. Changing a slider invalidates the previous draft preview; listen again before confirming.

The timeline editor requires Starter or above. Check the app for current feature access and purchase options before planning your edit.