Skip to main content
Version: Next

Add sound effects to a finished video

Goal: take a video that is already edited and get a new MP4 with short sound effects on the moments where something happens on screen: a pop where something appears, a whoosh on a move or a scene change, a shutter on a camera flash. Nothing is cut, and no music is added. The voice stays in front; the effects sit under it.

Prerequisites​

  • The video uploaded to the project (Files tab).
  • A chat AI provider, for the step that chooses the sounds.
  • A vision AI provider is optional and makes the result better. With one, step 1 also looks at the picture at each moment and notes what it shows ("a price tag pops up", "a man walks past"), so sounds fit what happens and ordinary filmed footage gets none. Without one the recipe still runs.
  • Your own sound files are optional. ReelBolt offers its built-in sound library (pops, clicks, whooshes, impacts, a camera shutter and more, free to use). If the project also has sound files of your own, they are offered too.

No speech-to-text provider is needed.

The recipe​

Add the Sound effects for your video template (video-sound-effects) from New workflow. It has three steps:

#LabelWhat it does
1Find the moments in the videoLists what happens in the picture: scene changes, things that appear, move or flash, text that comes in. Select your video as the source. Finding the moments uses no AI model. If your workspace has a vision model, it then looks at some of the moments to say what they show.
2Choose a sound for each momentThe sound designer picks a fitting sound for the moments worth one.
3Mix the sound effects inPuts each sound on its moment and gives you the new video.

In your brief you can say what the sounds should feel like, for example "soft and subtle" or "only on the big moments".

Under More options on the app page:

OptionWhat it does
Sounds on filmed footage tooOff by default: sounds go on designed moments (graphics, text, transitions), not on plain cuts between filmed shots or on people moving. Turn it on to get sounds there as well. A camera flash and an impact always get one.
Model that looks at the pictureWhich vision model looks at the moments. Leave it on the recommended one unless you have a reason.
Most looks at the pictureThe most moments the model looks at in step 1 (0 to 40). Leave it empty and the length of the video decides: about one look per six seconds, between 12 and 30. Set it to 0 to switch the looks off: no picture is sent to the vision model. When looks are on, the sound designer in step 2 may ask for up to 6 more looks at single moments. Each look is one small call to the vision model; the estimate you see before the run includes them.

What you get​

  • The same video, same length and same picture, with the sound effects mixed in.
  • The result of step 3 lists every sound and the moment it was put on, and says why a planned sound was left out when one was.

What it will not do​

  • It does not put a sound on every word of running captions. Words of captions that pop in one after another are recognised and left alone.
  • It does not cut, trim or reorder your video.
  • It does not make new sounds. It chooses from the built-in library and from your own files.

If something looks wrong​

  • Too many or too few sounds. Say so in the brief ("fewer sounds", "a sound on every scene change") and run again.
  • A sound is on the wrong kind of moment. The moments are found from the picture alone, so a pop-in on top of busy footage can be read as a plain change. The sounds are mixed into the new video, so a single sound cannot be moved or removed afterwards. Run the recipe again and say in the brief what to change ("no sound on the logo at the start", "nothing on the filmed shots"), or switch Sounds on filmed footage too on or off under More options. Running again makes a new video from your original; nothing is lost.
  • You want to place every sound by hand. Open your original video in the editor, show the Events lane and use Add sound here on a moment. That places sound files of your own; the built-in library is used by this recipe only.
  • No sound at all. The result of step 2 says why: for example the video has no clear moments (a plain talking video), or no chat AI provider is set up.