Voice-over, music and sound effects for your AI video
How to write a script that sounds natural, choose between ElevenLabs and Gemini voices, brief a music model, add sound effects and mix it all under a video made in FlexPic.
A silent clip looks unfinished. Good sound is often what makes a short AI video feel professional: a clear voice, a music bed that supports the mood and one or two sound effects at the right moment. FlexPic has three audio tools for this: Create voice-over, Create Music and Create Sound Effects. This article shows how to use them together.
Decide who makes the sound
Some video models generate audio with the picture, and the Audio switch in the video tools turns it on or off. Built-in audio is convenient for ambient sound and simple dialogue. For anything where the words, the music or the timing matter, switch it off and build the track yourself:
- You control the exact script.
- You can reuse the same voice across many clips.
- You can change the music without regenerating video.
Turning audio off can also lower the price of the video, as it does for Kling 3.0.
Voice-over
Write for the ear
A script for speaking is not the same as a script for reading.
- Short sentences. One idea each.
- Contractions. "It's" sounds more natural than "it is" in conversation.
- Numbers and symbols as words when pronunciation matters: "twenty percent", not "20%".
- Punctuation is your pacing control. A full stop gives a pause, a comma a short breath, an ellipsis a hesitation. If a line sounds rushed, add a full stop or split it.
- Budget the time. Spoken English runs at roughly 2.5 words per second, so a 15 second video holds about 35 words with room to breathe.
A sample for a 15 second video:
Mornings are hard. Your coffee shouldn't be. Brew in under a minute, anywhere you go.Choose a model and voice
In Create voice-over, the Audio Model list includes Eleven v3 and Eleven Multilingual v2 from ElevenLabs, and Gemini TTS from Google. The Voice list shows each voice with tags such as accent, gender and tone, for example Rachel (american, female, calm) or Daniel (british, male, middle aged). With Gemini TTS, the voice you pick is mapped to the closest of Gemini's own voices.
A practical process:
- Generate your first sentence only, with two or three voices.
- Choose the voice that matches the brand, not the one that sounds most impressive.
- Generate the full script with that voice.
Prices depend on the length of the text. A 500 character script costs about 50 credits with Eleven v3 and about 38 with Gemini TTS. Test one sentence before you generate a long script.
For another voice you own, FlexPic also has Voice Clone and Voice Changer tools. Use them only for your own voice or for a voice whose owner has clearly agreed. Cloning a real person without permission is prohibited by the Terms of Service.
Music
Brief the model like a composer
Describe the track with mood, instruments, tempo and role. Avoid naming artists or songs.
Mellow lo-fi beat with dusty vinyl crackle, soft piano chords and a slow laid-back groove, warm and relaxed, instrumentalIn Create Music:
- Quick Clip makes a 30 second piece. It is the right length for most social videos.
- Full Song makes a track of up to about three minutes and needs the Eleven Music model. Choose it in the Music Model row.
- Instrumental removes vocals. Switch it on for any music that sits under speech.
- Song Controls (BPM, Intensity, Key) are optional and set to Auto by default. Fix a BPM only when you must match an edit rhythm.
A Lyria 3 Clip track of 30 seconds costs about 50 credits. A quick clip from Eleven Music costs about 188, and a full song about 563. Generate the cheap clip first to settle the mood.
Fit the music to the video
Your video is probably shorter than your music. Trim the end of the track and fade it out over the last second, rather than letting it cut off. Choose a track with an even energy level for ads and one with a build for story videos.
Sound effects
Create Sound Effects turns a description into a short sound. Use Describe the sound and keep it concrete:
A heavy wooden door creaking open slowly in an empty stone hallwayUnder Sound Control, set a Duration (from half a second to 22 seconds, or leave it on Auto) and a Prompt Influence between Loose and Strict. The Loop switch makes a sound that repeats smoothly, useful for ambience such as rain. A 5 second effect costs about 13 credits.
Use effects sparingly. One well-placed sound, such as a whoosh on a cut or a click on a product reveal, does more than a constant layer.
Mixing everything together
Download the files and combine them in a video editor:
- Voice first. Place the voice-over, trim silences and leave a quarter second before it starts.
- Music under it. Set the music clearly lower than the voice. Raise it slightly where there is no speech.
- Effects on key moments. Line them up with the visual action.
- Fade the end. A one second fade on music and ambience avoids a hard stop.
- Listen on a phone speaker. Most viewers do.
Rights and disclosure
Generated audio is yours to use to the extent allowed by the underlying providers' terms and by law, as the Terms of Service describe. Do not imitate a real person's voice without permission. If a platform asks you to label synthetic voices or media, label them. The responsible use guide explains the details.
Next step
Put it all together in a complete advert with the 15-second product ad workflow, or see how the audio tools are priced in How credits work.