MMAudio

Available

zsxkib · zsxkib/mmaudio

Open in workspace

Tasks using this model

Video to Audio

Input contract

Required inputs: —
Optional inputs: seed, image, video, prompt, duration, num_steps, cfg_strength, negative_prompt
Output: video (uri)
Submission: Pinned community version

Provider billing

$0.000975 per hardware second (L40S, model page 2026-10-06).

Provider-side unit, for reference only. This site charges a fixed credit tier per execution — the price shown at quoting time is the price paid.

Contract notes

Schema also accepts image, but this product only submits video for the video-to-audio task. Prompt optional (defaults to ""); num_steps 25 / cfg 4.5 / negative_prompt "music" pinned server-side (adapter-exact defaults). MEASURED 2026-10-06 (batch E runs report): duration input is an output CAP — 3 s video + duration 8 → 3.000 s output; 10 s video + duration 8 → 8.042 s output; the adapter therefore pins duration=30 so output = source length. Hardware time scales ≈ linearly with output seconds (2.62 s predict for 3 s, 5.61 s for 8 s at $0.000975/s ≈ $0.0008 per output second) — the task prices probed source seconds per output second. Enabled 2026-10-06: batch E deployed-app E2E green (batch-e-e2e report).

Status

This model is catalogued and its contract is locked, but real executions have not passed acceptance yet. Nothing on this page should be read as a claim that it is usable today.

License: Community model; wrapper and underlying weights license not yet reviewed.
Last verified: 2026-10-05

Sources

All models