Every video needs sound design, and stock libraries only take you so far: the exact whoosh, the right rain, a door that sounds like your door. Generative audio changes the workflow — describe the sound, and the model makes it.
How text-to-SFX works
Sound-effect models are trained on huge collections of labeled audio. Given a prompt like 'heavy wooden door creaking open slowly', the model synthesizes a new waveform matching that description — original audio, not a library clip, so there are no duplicate-content collisions with other creators.
Prompt patterns that produce great effects
- Name the source and the action: 'glass bottle rolling on concrete' beats 'rolling sound'.
- Add texture words: hollow, metallic, distant, muffled, crisp.
- Describe the space: 'in a large empty hall' or 'small padded room' changes everything.
- State duration and shape: 'short impact', 'slow rise then fade'.
The prompt is the microphone placement: describe what the listener should feel standing in the room.
Library plus generation
SonicVox ships a curated, ready-to-use effects catalog for the common cases and a generator for everything else. Both feed the same editor as your voiceover and music, so scoring a scene stays in one place.
Sound is half the picture. When creators stop settling for close enough, their videos feel produced rather than assembled — and that difference is audible in the first five seconds.
