In this article
Keep music under voiceover low enough for every word to remain understandable, then shape its rises around complete phrases. Start with a suitable music bed, establish the voice on its own and review the combined mix by listening. An automatic ducking curve is a starting point for that review.
This workflow applies to recorded or AI-generated narration in an external audio or video editor. It does not assume that Tong supplies an automatic mixing control, and it offers no universal volume preset.
Choose music that leaves space for words
Before automating anything, play the most important sentence over your proposed music. If a busy melody, vocal sample or sharp percussion keeps pulling attention away, try a simpler arrangement or a different section of the track. A suitable bed can save more work than increasingly deep volume dips.
Keep the voice and music on separate tracks. Save an untouched copy of each, and mark the sentence that carries the product name or call to action. That sentence becomes your first listening checkpoint, rather than the loudest moment in the song.
Establish the voice before adding the bed
Listen to narration alone at a comfortable, consistent playback volume. Check quiet endings, breathy words and transitions between takes. Repair an uneven phrase before balancing the music; otherwise you may make the whole bed unnecessarily quiet to accommodate one weak word.
Bring in the music from a low level and raise it gradually while replaying that checkpoint. Stop where the mood is present and the sentence is effortless to follow. This is an editorial listening decision, not a measured loudness target. Check the receiving platform's delivery requirements separately.
Shape dips around complete phrases
Adobe's Premiere ducking guide documents audio-type tagging, duck amount, fade duration and fade position. It also explains that generated keyframes can be edited and that generating them again overwrites manual changes.
If using that workflow, identify speech as Dialogue and the bed as Music, then review the resulting automation. Listen for a music surge between two halves of the same sentence. Where that surge distracts, keep the bed lower across the short gap and let it recover at a meaningful pause. Preserve a copy before regenerating automation that you have adjusted by hand.
Try a simple product-film cue sheet
For a fictional 20-second product film, plan three musical moments. During the opening reveal, allow the bed to establish the mood. As the narrator introduces the product and explains its benefit, lower the music across the full thought. After the final spoken invitation, let it rise toward the closing image.