Add Sound18 October 2026

How to add sound to an AI video

Voice, music and sound effects, and the models that watch your clip. One clip with four soundtracks, with what each cost.

By Anand Murugan, who makes Chitra Studio

Some AI video models give you a silent clip. Others add their own sound, which may or may not suit the scene you have in mind.

Add Sound lets you choose the sound yourself: effects, music or a voice-over, made for the clip or taken from your library. Here is one clip with four different soundtracks, and what each cost.

Three kinds of sound

Chitra Studio makes three kinds of sound, each with its own models. In the prompt bar's audio mode they are three tabs:

Kind	What it is for	Models
Voice	Speech: voice-overs, narration, dialogue, in many languages	ElevenLabs, Gemini, MiniMax, Sarvam, Fish Audio and others
Music	Songs with vocals, instrumentals, scores and loops	Lyria, ElevenLabs, MiniMax, Mureka and Stable Audio
Sound Effect	Single sounds and ambiences: a bark, rain, a street	Stable Audio, ElevenLabs, Mirelo, MMAudio and CassetteAI

Each makes a sound on its own, which lands in your library. They are compared in AI voices compared, AI music generators compared and One sound effect, five AI models. Any of them can then be laid under a clip.

Adding sound to a clip

Add Sound in Chitra Studio. While the model works, a waveform rises over the clip; then the clip plays with its new sound.

Open a clip and choose Add Sound. There are two ways to give it sound:

From your library. Choose any voice, music or effect you already have, including a voice-over made on the Voice tab. It is laid under the clip on your Mac, free. You can keep the clip's own sound underneath, quieter, or replace it.
Make with AI. Choose Sound Effects or Music, then a model. The price is shown before you press Make.

The models differ in one important way: whether they watch the clip.

Model	Makes	Watches the clip	Good for
Mirelo SFX 2.0	Effects	Yes	Effects that follow the action, keeping any speech; two takes to choose from
Kling Video to Audio	Effects and music	Yes	Both together for one price, for clips of 3 to 20 seconds
Hunyuan Video Foley	Effects	Yes	A sound for each movement on screen, up to a minute
Sonilo Video to Sound Effects	Effects	Yes	Effects timed to each scene, keeping speech
MMAudio v2	Effects	Yes	Quick, low-cost drafts
Sonilo Video to Music	Music	Yes	A score timed to the clip's cuts and motion
Stable Audio 3	Music	No	An instrumental made exactly the clip's length
ElevenLabs Music	Music	No	The richest sound; at least ten seconds, cut to the clip

A description is optional. A model that watches the clip can work from the picture alone, or you can tell it what to listen for.
Improve fills it out. Press Improve to turn a few words into a full description, with Undo if you prefer your own. For a model that cannot see the clip, Improve reads the prompt the clip was made from, so the music is written for what is on screen.
Keep the original sound, or replace it. A sound from your library can sit on top of the clip's own sound, which is kept underneath and quieter, or replace it. A sound made with AI replaces it.
Your original clip is never changed. The result is a new copy of the clip with the new sound, and the sound itself is kept in your library too.

One clip, four soundtracks

The clip is five seconds from MiniMax H3 Max Turbo, which made its own sound along with the picture. Here it is with three other soundtracks.

The sound the video model madeMiniMax H3 Max Turbo, with the clipThe starting point.

Kling Video to Audio5 credits · 54 s · no promptIt watched the clip and made sound to match, with nothing typed.

Effects from MMAudio v21 credit · 27 sIt watched the clip, and followed a description of what to listen for.Effects prompt (MMAudio v2)"Heavy combat boots walking on gravel, tactical gear rattling, assault rifles cocking, magazines loading, weapon handling, tense outdoor wind ambiance."

Effects from MMAudio v2, music from Stable Audio 31 + 6 creditsTwo sounds made separately and laid under the clip together. MMAudio v2 watched the clip and made the effects: the boots, the gear, the wind, from the effects prompt beside this clip. Stable Audio 3 does not see the clip; it made a six-second thriller cue from its own prompt, below, and the cue sits under the effects, a little quieter.Music prompt (Stable Audio 3)"Cinematic modern thriller score, tense orchestral cue, rapid spiccato violins, surging cellos, sharp tremolo violas, driving double bass, suspenseful, urgent, 130 BPM, D minor, relentless sixteenth-note ostinato, intense dynamic crescendo building from quiet anxiety to a sudden sharp peak at 6 seconds, punchy transients, dry close-mic string textures, wide stereo field, cinematic compression"

Music made to the picture

Some models write music to a clip rather than sound effects. Sonilo Video to Music watched this dancer's four-second landing and scored it.

Sonilo Video to Music13 credits · 21 sPrompt"Cinematic thriller score with a tense, suspenseful mood. Features tremolo violins, driving cellos, and sharp ostinatos at a moderate tempo, gradually building in volume, dissonance, and rhythmic intensity toward a high-stakes crescendo."

What I took from it

No prompt is a fine place to start. A model that watches the clip, like Kling Video to Audio, gives you a matching first draft for 5 credits.
A prompt gets you specific sounds. Naming the boots, the gear and the wind told MMAudio v2 what to listen for, for 1 credit.
Layer effects and music. Sound and score made separately, then laid together, give you control over each.
Nothing is lost. Each try is a new copy of the clip, so you can keep all four and choose later.

Give your own clips sound

Chitra Studio is a Mac app for making and organising AI images, video and sound with more than 60 models. Every price is shown before you generate.

Download for Mac14-day free trial · 50 free credits · no card