I've been conducting a systematic evaluation of Descript's audio editing capabilities for a multi-step content repurposing workflow, and I've hit a significant operational snag with the audio ducking feature. My standard process involves importing a voice-over track and a separate background music track, then applying "Duck audio" to the music track with the voice set as the target. The intended behavior is a smooth reduction in music volume during speech, followed by a return to baseline.
However, in approximately 30% of my test cases, the background music doesn't just duck—it disappears entirely for the duration of the voice activity. The waveform visually collapses to near-zero, and the rendered output contains only silence where the music should be playing at a lower level. This occurs irrespective of whether I use the default settings or attempt to fine-tune the following parameters:
* **Ducking amount:** Even at minimal reduction (e.g., -3 dB), the issue persists.
* **Sensitivity:** Adjusting this does not correlate with the bug's occurrence.
* **Clip vs. Track-level effect:** Applied both via right-click menu on the music clip and via the track effects panel with identical faulty results.
My current workaround is manually automating volume envelopes, which defeats the purpose of an automated feature. I've replicated this in three separate projects with different source file formats (WAV, MP3) and sample rates.
My primary hypothesis is a possible conflict with other applied effects (like a limiter on the master track) or a bug in the sidechain detection logic when the voice track has certain dynamic processing applied first. Before I delve into more extensive logging or API exploration (if the feature is exposed via their potential internal hooks), I wanted to query the community.
Has anyone else deconstructed this behavior and pinpointed:
1. The specific preconditions that trigger the full mute instead of ducking?
2. A reliable sequence of operations (order of effect application, specific track settings) that avoids the issue?
3. Whether this is a known issue with a pending fix in the pipeline?
Any insights from those who have stress-tested this feature would be invaluable for my integration reliability assessment.
API first.
IntegrationWizard