Wavemaker treats sound as three distinct layers, each produced the right way.
On-screen dialogue (baked into the clip)
When a character speaks on camera, the dialogue is generated natively inside the video clip — real lip-sync, natural delivery, ambient room tone. This is not a dub laid over silent footage; the model performs the line.
- Lines are spoken verbatim from the script (your pasted dialogue is never paraphrased).
- Clip length is automatically fitted to the line — a long line gets the seconds it needs, with a natural breath after the last word instead of a hard cut.
- Delivery direction is honored: “she says, exhausted:” shapes the performance.
- Each character’s voice stays consistent across scenes via per-character voice signatures.
Off-screen narration
Narration is recorded separately with a professional-grade TTS voice and laid over the visuals with per-scene timing:
- Voice design. The narrator’s voice is designed to fit the piece — gender, age, accent, texture, pace, energy. Ask for anything: “a dry British documentary narrator”, “a warm Southern grandmother”.
- First-person pieces stay one voice. If a single on-screen presenter also narrates (“I built this company…”), the off-screen voice matches their on-screen voice — no mystery second narrator.
- Copy fits the cut. Narration is timed to each scene’s real duration at the voice’s measured pace, so lines never overrun their scenes or get cut off at the end.
- Verbatim lock. Pasted narration is used word-for-word, and the video is sized to fit it.
Music
- Generated score — written to the video’s mood, arc, and length; ends with the video instead of looping awkwardly.
- Your own track — upload a song and it is the score. Wavemaker analyzes the real BPM, structure, and lyrics, then cuts the video to the music. See Uploads & assets.
- Music direction — name a style in the prompt (“a driving synthwave track”) and the score follows it.
- Automatic ducking — music dips under dialogue and narration, beat-aligned.
Captions
Vertical/social formats add word-synced captions by default (each word appears as it’s spoken, in one of several animation styles). Any video can add them on request: “add karaoke-style captions.”
Silence is a valid choice
Say “wordless” or “no narration” and no narrator is added. Scenes carry diegetic sound — ambience and effects — with music doing the emotional work. A “silent film” style suppresses speech entirely.
Fixing audio after the fact
All of it is refinable in chat:
“Re-voice the narration with a deeper voice.” “Make the music 20% quieter under scene 2.” “Change Maya’s line in scene 4 to ‘We open at dawn.’”
Re-voicing a line or swapping music does not regenerate your visuals — only the audio layer changes.