Add Audio to Video Without Uploading It

The video and the audio are processed by your own browser, not on a remote server, so nothing is uploaded to Sonata AI. The only download is the processing engine itself, about 30 MB of WebAssembly fetched once from a public CDN and then kept in the browser cache.

Replace drops the original sound completely and leaves only the new track, which suits silent AI clips or noisy footage. Mix keeps what the camera recorded and lays the new track on top, so voices, laughter or room tone stay audible under the music.

Each source gets its own level. At 100% nothing changes; lower values push a sound into the background, higher ones lift a quiet recording. Moving them separately is what lets a voice and a song sit together instead of fighting for attention.

The finished file is exactly as long as the video. A song that runs longer is cut where the picture ends. A short jingle can be looped to the last frame, or left to stop, after which you get silence in replace mode or the original sound alone in mix mode.

When the video is H.264 inside an MP4 or MOV file, the frames are copied as they are and only the sound is rebuilt. That is quicker and costs no quality. WebM files and other codecs are converted to H.264 instead, which takes longer.

Cutting a track at the last frame can sound abrupt, especially in the middle of a chorus. Switch on the optional fade-out and the audio eases away at the end of the clip rather than stopping dead.

There is nothing to sign up for and no credits to spend, and the MP4 you download carries no logo. A progress bar shows how far the job has got, and you can cancel at any point if you spot a wrong setting.

Text-to-video and image-to-video models often deliver mute clips. Put music or a short voice line under the result and it is ready for a post or a presentation, without opening a full editing suite.

Record the screen first and the narration afterwards, once you know exactly what to say. Replace mode puts your voice over the capture; mix mode keeps clicks and notification sounds faintly in the background.

Wind, traffic or a noisy crowd can spoil footage that otherwise looks good. Replace the noise with music, or mix with the original turned well down to keep a hint of the atmosphere.

A short background track gives a demo or a teaser some energy. Keep the music low if a voice explains the product, and use the fade-out so the clip ends cleanly instead of mid-beat.

Add a narration you recorded as an M4A voice memo to a video exported from a slideshow app, and share the result as an ordinary MP4 that relatives can open without special software.

If you have recorded a translated narration, replace the original voice with it and keep one video file per language, all sharing the same picture.

If people talk in the clip, leave the original near 100% and bring the new track well down; somewhere between a quarter and a half of full level is a sensible starting point. Then adjust until every word is still easy to follow.

The 200% ceiling helps with a quiet phone recording, but a track that is already loud can distort when pushed further. If you hear crackling on the peaks, lower that slider instead of raising the other one.

A song slightly longer than the video, combined with the fade-out, gives the cleanest ending. Looping works best with material made to repeat, such as ambient beds or simple beats; a song with a clear intro and outro will reveal the seam.

Many AI-generated videos and screen captures have no sound at all, or only a faint hiss. Replace gives you a clean slate, whereas mix would keep that hiss running quietly under your music.

If your phone or editor offers H.264 MP4 as an export option, use it. The tool can then copy the picture without re-encoding, which keeps it exactly as it was and finishes sooner. Anything else has to be re-encoded first.

Check the result on headphones and on a phone speaker. A music level that sounds fine on one can bury a voice on the other, and small speakers are where most short videos end up being played.