Why did Normalize change the balance between my product-ad voiceover and music?

Identify What Normalize Actually Changes

Version scope: the track-by-track versus shared-gain comparison here follows the cited Audacity Development Manual. The newer web manual describes selection peak normalization and stereo-pair controls but does not spell out the same multi-track comparison. Do not assume every installed release exposes identical controls or scope; verify that version before applying this comparison.

Audacity's Normalize effect can set a track's peak amplitude to a specified target. When several tracks are selected, each track is normalized independently to that target. The tracks do not necessarily receive the same gain adjustment.

Illustrative exercise: assume two fictional tracks with no DC offset and no other processing, with narration peaking at -6 dBFS and music at -18 dBFS. Independent normalization to -3 dBFS gives illustrative gain changes of +3 dB and +15 dB. These teaching values are not an advertisement delivery specification.

This does not prove that the two tracks will sound equally loud. Peak amplitude is not perceived loudness, and musical dynamics differ from speech dynamics. Nevertheless, changing the relative peaks can substantially change the mix.

The Development Manual explicitly distinguishes independent peak normalization from applying an identical gain to all selected tracks.

The Development Manual's Normalize page explicitly compares independent track peaks with shared amplification.

For the separate listening question, see the guide to peak level versus perceived loudness.

Use Amplify When the Existing Relationship Must Remain

Audacity's Amplify effect applies the same gain amount to all selected tracks or channels. If narration and music already have an acceptable relationship, a common gain adjustment preserves their relative peak difference.

For that same fictional gain-only comparison, adding +3 dB to both tracks gives peaks of -3 dBFS and -15 dBFS. The original 12 dB peak difference remains. This demonstrates shared gain rather than independent peak targets; it does not establish equal perceived loudness or the peak of a combined mix.

Check the Amplification and New Peak Amplitude fields before applying the effect. Do not enable clipping merely to achieve a louder result. A shared gain operation cannot fix an already incorrect voice-to-music balance.

The Amplify manual states that all selected tracks and channels receive the same gain amount.

Illustrative Exercise: Compare Two Inputs

Illustrative exercise — not a measured advertisement or software test. Use the fictional gain-only inputs calculated above, with no DC offset or other level processing. On separate project copies, compare Normalize to -3 dBFS with Amplify by +3 dB.

Record actual individual peaks and the combined listening result rather than copying the calculated values into an observation field. Inspect the independent stereo-channel option as well: its default paired-channel setting preserves the existing left-right relationship.

Verify the Mix Before Delivery

Save a working copy before processing. Check that the intended tracks and time ranges are selected, then compare the original and processed versions at matched monitoring volume.

Listen specifically to product names, prices, sizes, discount conditions, and the final call to action. Check the exported mix as well as the individual tracks. A satisfactory individual peak does not guarantee that the combined mix avoids clipping.

If the narration is difficult to understand, do not automatically normalize both tracks again. Revisit the actual mix balance. Audacity also documents DC-offset removal under Normalize, but that function should not be confused with balancing narration against music.

The multi-track comparison follows the cited Development Manual pages for Normalize and Amplify. The current Normalize web manual separately supports the stereo-pair and DC-offset controls but does not spell out that multi-track comparison. Check the installed version and processing scope; the different wording alone does not establish that behavior changed.

Illustrative Exercise — Independent Peaks Versus Shared Gain

These are illustrative exercise rows. Save a copy and record your own observations; the example is not an executed acceptance test.

Download review worksheet (CSV)