If Sound Finder misses audio when creating labels, check the selected range, detection threshold and grouping settings. In current Audacity, use Label Sounds, which replaced Sound Finder. Test a short passage before processing the full track.

This guide is for a Mac user who wants a separate region label around each intended sound, such as synthetic tones separated by quiet gaps. The goal is to recover missing labels and incomplete boundaries without changing the audio.

Find the right labeling tool for your Audacity version

Audacity’s retained documentation says Sound Finder has been superseded by Label Sounds. The older analyzer may remain available in installations carried forward from an earlier version. Its absence from the menu does not establish that your installation is broken. Sound Finder — Audacity Manual

The official Mac download page currently lists Audacity 4.0.1 and marks the separate ARM64 download as having no plugin support. Keep that build limitation in mind when investigating unavailable analyzers; do not assume installing an old plugin will resolve it. Download for macOS

With audio selected, look for Analyze > Label sounds. If it is unavailable, check Analyze > Plugin manager for the analyzer and enable it if listed. The official manual explains that the analyzer menu depends on installed plugins. Label Sounds, Analyze

Menu wording also differs between generations. The current manual uses Edit > Label > Add label for manual labels; the retained older manual uses Edit > Labels > Add Label at Selection. Follow the wording in your installed version. Edit, Label Tracks

Check whether audio is missing or only its label is missing

First, play the passage that appears to have been skipped. If the sound plays but has no separate label, continue with detection troubleshooting. If it does not play, label settings cannot supply the missing audio: analyzers inspect existing audio without changing it, and label tracks contain markers rather than audio. Analyzers, Label Track Item

Next, inspect the selected time range. Analyzers examine only the selected audio. A passage outside that range cannot receive a label from that run. Include the missed sound and the gaps on both sides so that you can judge its boundaries. Analyzers

To select a test range in the current interface, click inside the waveform and drag horizontally. Keep the drag within the intended audio track: dragging vertically through other tracks includes them too. Avoid dragging the clip header, which moves the clip. Selecting and moving

For an older Sound Finder installation, select only the audio track, leaving existing label tracks outside the selection. Its documentation recommends this to produce a fresh label track and avoid mixing new results with existing labels. Sound Finder — Audacity Manual

Diagnose the result before changing settings

A missing separate label can have several explanations. Use this sequence as a diagnostic aid derived from Audacity’s documented controls. Change one setting per trial and compare the same passage each time. Label Sounds, Label Sounds — retained manual

What you observe First thing to investigate
A quiet sound falls outside all labeled regions Detection threshold
Several intended sounds share one region Silence duration, then label interval
One intended sound becomes several regions Short internal pauses
A label starts late or ends early Threshold, then boundary padding
Labels cover the gaps Label type

Recover a quiet sound with the threshold

Threshold level (dB) determines what counts as sound. The retained manual warns that a threshold set too high can miss beginnings and endings; one set too low can classify the background as continuous sound. Lowering the threshold means choosing a more negative value. Label Sounds — retained manual

For an illustrative test, imagine three synthetic tones with peak levels of −18, −38 and −18 dB, separated by digital silence. With Peak level measurement and a −30 dB threshold, the middle tone falls below the threshold. A trial at −45 dB puts all three above it. These are invented inputs demonstrating the comparison, not measured results or a universal preset.

Try the threshold adjustment on your short selection first. Record whether the quiet sound gains a region and whether unwanted material also appears. Keep the result only if it improves both coverage and separation.

Separate sounds that have been grouped together

Minimum silence duration decides whether a quiet gap separates sounds. Minimum label interval groups labels that would otherwise be too close together. Check them in that order: first establish a boundary, then check whether grouping combines the resulting sounds. Label Sounds

Consider two synthetic tones, each two seconds long, separated by 0.4 seconds of digital silence. As an illustrative trial, use Minimum silence duration of 0.2 seconds and Minimum label interval of 0.1 seconds. This deliberately tests a gap shorter than the documented one-second default without adding a large grouping interval.

Use different values when the intended task requires them. If several brief tones should form one item, your desired output is a single region. Write down the expected grouping before tuning; otherwise, a larger label count can look like progress even when it creates unnecessary editing work.

Restore incomplete beginnings and endings

If a detected region excludes a quiet attack or fading tail, revisit the threshold first. Then consider Maximum leading silence and Maximum trailing silence to include a margin around the detected sound. The retained manual documents limits on those margins where neighboring sounds leave insufficient room. Label Sounds — retained manual

For example, an illustrative 0.05-second leading margin and 0.10-second trailing margin may provide useful room for reviewing a synthetic tone’s edges. Treat those values as trial padding. Listen across each boundary rather than assuming the extra time includes everything you intended.

Run a controlled labeling pass

Use the following procedure once you know which symptom you are correcting:

  1. Select a short audio range containing a correctly detected sound, a missed sound and their surrounding gaps.
  2. Open Analyze > Label sounds.
  3. Choose Region around sounds so the result spans the audio rather than marking positions or gaps.
  4. Keep the measurement method fixed while testing one threshold or duration change.
  5. Apply the analyzer and inspect the result before expanding the selection. Label Sounds

Keep a brief trial note, such as “quiet middle tone recovered; first and third regions still separate.” Before another trial, undo the previous analyzer result if you do not need it for comparison. Audacity documents that adding the analyzer’s label track can be undone. Analyzers

Changing measurement method is a separate experiment. The current documentation suggests RMS level when isolated sharp transients in gaps trigger detection. Recheck the threshold after switching methods; do not assume a value chosen for peak measurement behaves identically. Label Sounds

Finish exceptions manually and check the boundaries

When the desired division cannot be expressed by levels and quiet gaps, use manual labels for the remaining exceptions. In the current interface, stop playback, select the intended time range and choose Edit > Label > Add label, then type its name. Manage labels opens the editor for reviewing the project’s labels together. Edit

To correct an existing region in the current interface, drag either endpoint. Double-click its name to rename it. Dragging the name tag moves the label, so use an endpoint when only the start or end needs adjustment. Label Track Item

Preserve the documented limitations of older installations: the retained Label Sounds manual specifies 10-millisecond processing chunks and a maximum of 10,000 labels per run. It recommends processing sections for larger jobs. Those are documented boundaries of that analyzer, not a promise of sample-accurate automatic placement in every version. Label Sounds — retained manual

Finish by listening across the quietest sound, the shortest intended gap and a representative start and end. Accept the labeling pass when those checks match your intended regions; correct isolated exceptions manually before using the labels for subsequent editing.

Sources