MMAudio: 4 Video-to-Audio Before/After Tests

MMAudio generates synchronized audio for a video. This post runs four before-and-after tests and shows the resulting clips with audio.

Model

MMAudio on Wiro

4 before/after tests

Test 1

Prompt

ocean waves crashing on the beach, seagulls, wind

Before (input video)

After (video with generated audio)

Test 2

Prompt

horse galloping, hooves on dirt, snorting, breath

Before (input video)

After (video with generated audio)

Test 3

Prompt

lightning storm, thunder, heavy rain, strong wind

Before (input video)

After (video with generated audio)

Test 4

Prompt

busy city street, cars passing, distant siren, light rain

Before (input video)

After (video with generated audio)

What to watch for

• Use a concrete prompt. Name the main sound sources.
• Keep prompts short. Too many sounds can fight each other.
• Try a mismatch prompt on purpose to see how much the model follows text vs visuals.

Try it

Run MMAudio on Wiro: https://wiro.ai/models/wiro/mmaudio