Original Voice Removal in Auto Mixing

This article explains how original voice removal works when Auto Mixing is enabled in dubbing projects, and how timing boundaries and crossfade settings affect the final result.

Overview

When Auto Mixing is used, the system removes the original dialogue track and replaces it with background audio.
The accuracy of this process depends on the speech boundaries inherited from the Transcription task.

How speech boundaries are defined

Step 1. Transcription defines timings

Speech boundaries are taken from the original speech timings generated during the Transcription task.

In most cases, these timings are accurate.
However, they may occasionally contain issues, such as:

  • Missing dialogue lines

  • Segments extending beyond actual speech (e.g. breathing, laughter)

Step 2. Boundaries are used for voice removal

These boundaries define where the original voice should be removed during Auto Mixing.

To review them:

  1. Open the Lip-sync task.

  2. Enable Show voice removal boundaries on the timeline.

How original voice is removed

Step 3. Select the removal method

The system removes the original voice in one of the following ways:

  • M&E track replacement, if an M&E track is available

  • AI-based speech removal, if no M&E track is provided

Step 4. Apply crossfade smoothing

Voice removal is applied with crossfade smoothing.
This ensures a smooth transition between removed dialogue and background audio.

Crossfade smoothing:

  • Gradually fades out the original voice

  • Blends in the background audio

  • Reduces audible artifacts and abrupt cuts

Configure crossfade duration

Step 5. Adjust the crossfade parameter

You can control the fade duration using the Crossfade parameter.

To adjust it:

  1. Open the Lip-sync or Auto mixing task.

  2. Go to Sound → Advanced.

  3. Set the Crossfade value.

Recommended values

  • Default: 300 ms

  • Shorter values (100 ms or 50 ms) may work better depending on the content

How close segments are handled

Step 6. Understand short gaps

If the gap between two speech segments is less than twice the crossfade value, the audio between them will also be removed.

This behavior:

  • Prevents sudden volume changes

Result

Accurate speech boundaries combined with appropriate crossfade settings ensure:

  • Clean removal of the original voice

  • Smooth background transitions

  • Stable audio levels in the final mix