How to add music or replace audio in a video
Choose what happens to the original sound. Then line up the new track, listen and save.
Free toolAdd Audio to VideoLayer music, cut and position audio clips, change speed or replace a video’s sound. Preview and save locally with the picture preserved.Add audio to your videoReplace for a new soundtrack. Mix to keep both.
Replace removes the original audio from the result. Mix adds another track alongside it. Neither operation needs to change the picture when the video codec fits the output container.
| Your task | Start with | Check before saving |
|---|---|---|
| Add music under speech | Mix, original at its normal level, music quieter. | Every spoken word stays easy to hear. |
| Return cleaned sound to a video | Replace, full added level, no loop or fades. | The replacement has the same timing and duration. |
| Put a voiceover over ambient sound | Mix, original quieter than the voiceover. | The original speech does not compete with the new voice. |
| Give a silent clip a soundtrack | Replace or mix. There is no original sound to keep. | The audio covers the intended portion of the clip. |
Use a browser to place and cut your soundtrack
The add audio to video tool works locally with short SDR clips. Open your video and audio files. Choose Mix both or Replace, then place and balance the clips. Your mix lets you hear the rendered sound before saving.
- Drag each clip to its position. Move a long recording before zero to use a later section.
- Trim its edges, or open Trim source to choose Source in and Source out.
- Place the playhead and split a clip to rearrange its pieces. Overlap clips to layer sounds.
- Choose speed, volume and fades for each piece. Check both ends, then save a new file.
A full editor is a better fit for several picture clips, pitch-preserving speed changes or music that changes level during speech. Clideo’s official manual describes its placement, trimming, volume and loop controls. Its export instructions include a paid watermark-removal option.
Two start times answer two different questions
The placement time belongs to the video. The source position belongs to the audio. Keeping them separate avoids accidentally skipping the opening of your picture.
Start in video: 2 seconds
The new audio first becomes audible two seconds into the video. The picture still begins at zero.
Source in: 8 seconds
The audio’s first eight seconds are skipped. Its eight-second position is now heard at video second two.
Looping: on
Only the part after second eight repeats. The skipped intro stays out of every repetition.
Understand what happens at the end
At 1× video speed, the picture keeps its original duration. A short track either repeats or ends early. In replace mode, an early finish leaves silence. In mix mode, the original sound can continue.
| Example | Settings | Result on the video timeline |
|---|---|---|
| 12-second video, 5-second audio | Start 2, source in 1, play once. | Added sound plays from second 2 to 6. The last six seconds have no added audio. |
| The same two files | Start 2, source in 1, loop. | The four-second selection plays twice, then its first two seconds. The video still ends at 12. |
| 4-second video, 10-second audio | Start 0, source in 3, play once. | Audio seconds 3 to 7 accompany the full picture. |
| 3-second video, 8-second audio | Start −4, source in 0, play once. | Audio seconds 4 to 7 accompany the picture. Dragging left selects later content. |
These are worked timeline examples, not performance measurements. Fades affect only the added track. A fade out ends at the piece’s end. Trim an overhanging piece if the fade should end with the picture.
Change speed without losing the relationship between clips
Video speed changes the picture and its original sound together. A 12-second video at 2× becomes six seconds. Added audio has separate speed controls and keeps its chosen position on the finished timeline.
Audio at 2× plays in half the time with a higher pitch. At 0.5× it lasts twice as long with a lower pitch. Floi uses the same rendered audio for preview and export. It does not preserve pitch when changing speed.
Keep the balance without clipping
Combining tracks can create a peak louder than either source. Floi checks the rendered samples and applies one common gain when the peak exceeds 0.98. Your relative balance stays the same.
- Peak protection is sample-based. It is not a true-peak limiter or a loudness target.
- The AAC or Opus encode can change the final peaks slightly. Inspect the exported sound for a delivery specification.
- Use the audio loudness normaliser for LUFS and true-peak measurements.
Preserve the picture without promising lossless sound
Copying encoded video packets avoids another picture encode. Resolution is retained. At 1×, frame timing is retained too. Other video speeds divide each frame’s timestamp and duration by the selected rate. The container must support that video codec, and rotation metadata must remain representable.
The sound follows a different path. Mixing, fades and peak protection require decoded samples. Floi renders stereo at 48 kHz, then encodes AAC or Opus with a 192 kbps target. Audio quality and container metadata can change.
What we checked in the saved files
We inspected synthetic exports on 29 and 30 October 2026 using Chrome 156 and automated WebKit 27.2. The computer was a MacBook Pro with an M1 Max and 32 GB memory. FFprobe checked the video; FFmpeg decoded the saved sound.
| Test | Observed result in both engines |
|---|---|
| Three-second H.264 clip, 72 frames | MP4 retained all 72 encoded picture packets with matching hashes. Picture and container duration were 3.000 seconds. |
| Three-second VP9 clip, 72 frames | WebM retained all 72 picture packet hashes and a 3.000-second container duration. |
| Separate skip and placement | Known audio impulses at 0.3, 0.9 and 1.5 seconds were placed at 0.6, 1.2 and 1.8 seconds. Both saved formats matched those positions in the decoded samples. |
| Added audio formats | MP3, M4A, AAC, FLAC, Ogg Opus, WebM Opus and WAV at 44.1, 48 and 96 kHz passed impulse timing checks. |
| Variable frame timing and rotation | The variable-rate fixture retained 72 packets across 4.500 seconds. A separate rotated MP4 kept its 90-degree transformation. |
| Long audio moved before zero | Audio impulses at 4.3, 4.9 and 5.5 seconds were heard at 0.3, 0.9 and 1.5 seconds after a −4-second placement. |
| Overlapping music clips | Two equal source tones at 100% and 50% retained an amplitude ratio within 0.47 to 0.53 in the overlap. |
| Video at 2× and 0.5× | The three-second fixture saved at 1.500 and 6.000 seconds. All 72 picture packet hashes matched, with presentation times scaled to the chosen speed. |
| Five-minute fixture | A 320 × 180 clip at one frame per second saved with 300 matching picture packets. The result lasted 300.000 seconds. |
The impulse test used a 0.5-second placement and skipped 0.2 seconds of the source audio. The tool measures encoder delay with its own test signal before saving. These fixtures do not establish maximum file sizes or results on a physical phone.
Use FFmpeg for repeatable soundtrack replacement
FFmpeg can handle longer recordings, batches and codecs outside browser support. These commands assume a video whose picture codec is compatible with MP4. They create a new file.
Replace the sound and keep the full picture
Copy the primary video, encode the new audio as AAC, and pad short sound with silence. The shortest option then ends the result with the video.
ffmpeg -i input.mp4 -i soundtrack.wav -map 0:v:0 -map 1:a:0 -c:v copy -c:a aac -b:a 192k -af apad -shortest -movflags +faststart replaced.mp4Mix looping music under an existing soundtrack
Loop the music input, lower its level and mix for the original track’s duration. This example requires original audio. The limiter is separate from floi’s common-gain protection.
ffmpeg -i input.mp4 -stream_loop -1 -i music.wav -filter_complex "[1:a]volume=0.25[m];[0:a:0][m]amix=inputs=2:duration=first:normalize=0,alimiter=limit=0.98:level=false[a]" -map 0:v:0 -map "[a]" -c:v copy -c:a aac -b:a 192k -shortest -movflags +faststart mixed.mp4See FFmpeg’s stream-copy documentation and its amix and apad filters for the exact behaviour. Copying requires a compatible codec and container.
Repair the soundtrack, then put it back
Keep the video untouched while you work on its sound. Extract the original audio, repair it and replace the old track with the repaired file.
- Use Extract Audio from Video to save the soundtrack.
- Reduce noise with the noise remover, or measure and adjust loudness.
- Open the original picture and repaired audio in the soundtrack studio. Choose Replace, 100% added level, no loop and no fades.
- Check lip sync near both ends before saving.
Removing silence or changing speed breaks the one-to-one timing. Aligning the start cannot fix sections removed from the middle. A desktop timeline editor such as Shotcut is better for matching those picture edits and varying music levels over time.
Choose music you can publish
Bring your own recording or music licensed for your intended use. A tool cannot supply permission by adding it to a file. Check attribution requirements, commercial use and distribution limits with the source.
Music licensed inside a social app may have different terms from a song attached before upload. When using that library, add the sound in the destination app. Keep a clean master for other destinations.