Sound Effects V2

What Sound Effects V2 is good at — and what it's not

Sound Effects V2 is the model to reach for when you need a specific sound and don't want to dig through a library. It's a text-to-sound-effect model: describe what you want to hear — "heavy rain on a tin roof," "a sci-fi door whoosh," "distant city traffic at night" — and it generates the audio to match. That covers the whole non-musical palette: one-shot effects like impacts and UI clicks, foley like footsteps and cloth rustle, and ambient beds like room tone and crowd murmur.

The capability that makes it production-ready is seamless looping. Flip looping on and the clip is built to repeat without an audible seam, so a 10-second rain bed or machine hum can run under a scene of any length. Paired with a generation window of up to 22 seconds, you can author both the short hits and the long atmospheres a project needs from one tool.

The economics make it practical to use everywhere. Generating every cue in a project is cheaper than licensing a sound pack, and you get exactly the sound you described rather than the closest match in someone else's catalog. It pairs naturally with generated video: render a silent clip, then generate the footsteps, ambience, and impacts to sit under it.

Where it's the wrong tool: it does not make music — for full musical tracks, switch to Lyria 3 Pro — and it does not produce spoken voiceover, which is Multilingual V3's job. Sound Effects V2's lane is non-musical audio: effects, foley, and ambience.

A reasonable mental model: default to Sound Effects V2 for any non-musical sound — one-shots, foley, and loopable ambience. For music, use Lyria 3 Pro; for narration and dialogue, Multilingual V3.