Docs
plen

Music for video

This module adds music that is composed FOR your finished clip. The AI system watches your video, detects the tempo and the moments of change on its own, and composes three versions of a track exactly as long as the film. You play each version straight away on your own clip, with the picture, pick one, set the volume, and the bed drops under the voice whenever someone in the recording speaks. The finished file lands in your Asset Library and can go on to get a voiceover or captions.

What it's for

  • For whom: anyone preparing video for social media, a website or a presentation who wants a bed without digging through music libraries and without matching cuts by ear.
  • Use it when: you have a finished clip (your own, from the generator, or one that already has a voiceover) and you want the music to keep its rhythm.
  • If you want to add a voiceover or captions, use the neighbouring Add voiceover and Captions tabs in "Video editing" (Dubbing + Lipsync) - ideally ON the file with music, because the bed is then a separate track under the voiceover and ducks by itself.

Info

The music is written to the picture, not the other way round. The model sees your cuts and writes its tempo to them. It cannot, however, be handed a list of cuts - with a very dense edit it lands on the overall pulse and the strongest changes, not on every cut. That is why you get three versions to watch with the film, and the system suggests the one that sits best in the edit.

Before you start

  • Permissions: a role allowed to use services (member and up).
  • Module enabled: "Music for video" must be active for your company. If you do not see it under "Video editing", ask for it with the Request a module button on the dashboard or via Report in the top bar.
  • Material: a video file from disk (MP4, MOV, MKV or WebM, up to 200 MB, up to 10 minutes, at most 4K and 120 frames per second) or a clip from the Asset Library. The clip can be silent or have its own sound. One composition covers a clip up to the length set for your company (90 s by default).

How to add music

  1. In the company menu open Video editing and pick the Music tab (next to "Add voiceover" and "Captions").
  2. Point at a clip - upload a file or pick one from the library (a film that already has a voiceover works too).
  3. Pick a music style from the list or leave "AI decides" - the model then describes the mood of the picture itself. Some styles have a short sample to listen to. If your company has custom descriptions enabled, describe the music you want in your own words, in any language - AI rewrites it in the background, while you type, into 3-5 English sound words (genre, bass, drums) and shows under the field exactly what the composer will get. Words about the song's arc (build-up, drop, intro) are dropped - they impose their own rhythm and break the fit to the cuts.
  4. Click Compose 3 versions. This uses one generation from the module's limit. Composition usually takes 1-3 minutes; you can refresh the page, nothing is lost.
  5. Watch the three versions. Each plays on your clip, with the picture and the music underneath - the way the finished film will look. We pre-select the one that sits best in the edit; keep it, or pick the one you simply like. The volume sliders from the mix drive the previews too.
  6. Set the mix: music volume, and for a clip with its own sound also that sound's volume and the "Duck music under speech" slider - how far the music drops when someone speaks (0% = it plays evenly). All three sliders are set automatically from measuring the loudness of the clip and the chosen version; you can adjust them, and "Set automatically" brings the measured values back. We nudge the track to the cuts ourselves, by a few dozen milliseconds, invisible to the eye.
  7. Click Render the film with music. The picture stays untouched (copied losslessly), only the sound changes. The file lands in the Asset Library and in the library under the "Video editing" tabs - from there one click adds a voiceover or captions to it.

Good to know

  • Three versions, because the model is not deterministic - the same clip and style give different music every time. Three tries are the only way to have a choice.
  • The preview sounds like the file. The mix sliders, ducking under speech included, drive the previews live - computed the same way as in the rendered film. If no version sits in the edit, compose again - that is three new versions.
  • Ducking reacts to speech, not to the background. When composing we measure what plays all the time in your clip (engine, wind, street noise), and the music only drops when someone speaks louder than that. Under the mix you see in how many places speech was found, and the previews show a bar marking where the music ducks. If you mute the film's sound, there is no ducking - nothing to duck under.
  • The music does not jump between sentences. It drops just before the first word, stays down through pauses shorter than 2.5 s and only comes back gently after a longer silence.
  • The music is as long as the clip, so if you trim the film later in another module the track will be too long - compose it again then.
  • Order with a voiceover. Music first, then voiceover: the bed is then a separate track under the voice and ducks by itself. Voiceover first, then music also works - the bed then ducks under the voice in the recording.
  • What it costs. The limit counts generations (one composition = one slot) regardless of clip length. The mix itself does not use the limit - render as many times as you like.