Your AI video has no sound. Three free tools fix that
Claude makes good motion graphics and then you press play and hear nothing. Three free, open-source tools to find sound, generate it, and land every hit on the right frame.
Claude is genuinely good at motion graphics now. Then you press play and it sounds like a muted call. Visuals without sound read as a draft no matter how good the animation is, and most people stop there because audio feels like a separate craft. Three free tools got me past it.
The three pieces
- soundfetch finds free CC0 sound effects, so you are not paying for a library or risking a claim.
- earshot listens to your edit and generates new sounds in code when nothing existing fits.
- HyperFrames treats video as code, which is what lets a sound land on an exact frame instead of roughly near it.
The order that works
- Build the video first and get the timing right while it is silent. Sound cannot rescue bad pacing.
- List the moments that need a sound — each cut, each element that appears, each number that lands. Usually fewer than you expect.
- Run soundfetch for those moments and take what fits.
- Use earshot for anything soundfetch has no match for, which tends to be the abstract interface noises.
- Place each sound on its frame in HyperFrames, then watch the whole thing once with your eyes shut. If you cannot follow the edit by ear, the sound is decorative rather than structural.
The part people get wrong
A sound a few frames late reads as broken even when the viewer cannot say why. That is the whole reason to do this in code rather than by dragging clips around a timeline — a frame number is exact and a drag is not.
Second thing: most social video is watched muted. Sound should reward turning it on, not be required to follow what is happening.