How do you add voiceover and music to an AI animated short film?

Add voiceover and music as a deliberate audio pass after your clips are generated: script narration scene by scene, generate one consistent voice per character, brief music against your film's emotional beats, then mix — narration around -12 to -6 dB over a music bed at -20 to -25 dB — and run a review pass on the finished cut.

Script the voiceover scene by scene before you generate a single line of audio. If your film lives in the invideo agent — invideo is an agentic video creation tool with the current video and image models and audio tools in one project — it already holds your full script, character arcs, and shot breakdown, so ask it to draft narration lines and timing notes per scene rather than writing them cold. If you keep a production document, add your sound rules to it in one pass — one documented production wrote diegetic sound logic straight into asset briefs ("hard material, so it makes a horrible sound when it falls") so picture and audio stayed coherent from the design stage.

Then generate the voiceover itself. For narration, generate it as a separate audio track and keep one consistent voice per character across every scene — voice drift between scenes reads as badly as visual drift. For on-screen dialogue shots, newer video models such as Veo 3 and Kling can generate speech and ambient audio natively inside the clip; the invideo agent routes individual shots to these models, so lip-synced lines can come out of the generation step itself while narration gets laid over the top in the edit. All of these models run inside invideo, so you don't need a second platform for the dialogue path.

For music, brief it against your film's emotional structure rather than asking for a generic mood: if your script escalates across distinct emotional beats, the score should shift register at each beat boundary the same way the lighting does. Whatever the source — AI-generated or library — confirm the license covers publishing before you commit the track to the cut.

Mix in the edit. Documented productions assembled final cuts in Adobe Premiere Pro or DaVinci Resolve, and you can also finish inside invideo's editor. Keep narration around -12 to -6 dB and the music bed at -20 to -25 dB, ducking the music further whenever a character speaks so dialogue stays clear.

Finally, run an audio review pass on the rough cut. Upload the assembled cut back to the invideo agent with an open "what's working, what's not" prompt: in one documented ~90-second production ($870, 2 days, ~400 video generations), this pass flagged pacing problems, SFX issues, and a reveal shot playing at the wrong emotional register — a sound-and-timing error the director hadn't noticed. Fix what it flags, re-balance the levels, and export.

Watch some of these to see what works for you:

This is the step that most people skip, but it's actually extremely useful.

— the director of a documented AI horror short film, on running a rough-cut feedback pass before final export