A 14.8-second voice-over bills 15 seconds on H3 Max lip sync: $1.50

A 14.8-second voice-over on MiniMax H3 Max Lip Sync is the longest audio the route accepts, and it is billed as 15 seconds: 15 x $0.10 = $1.50 at 768p, $3.00 at 1080p, or 15 x $0.0625 = $0.9375 at 480p. The catalog describes the price as per ceil(audio second).

The window

Audio must be 5 to 14.8 seconds, Sume-hosted, and under 10 MB. The catalog explains why the upper bound is exact: the provider rejects shorter audio and silently clips longer audio, so Sume refuses a request outside the window rather than clamp it and drop speech. The output length follows the audio.

Why the rounding matters

A 14.8-second file and a 14.01-second file both cost the same, 15 seconds, so you lose nothing by using the full window. A 5.0-second file costs 5 x $0.10 = $0.50 at 768p. If your script runs to 16 seconds, you cannot trim it by a second with this route; split the voice-over into two takes of 8 seconds, each costing $0.80 at 768p, for $1.60 in total.

Checks before you submit

Measure the audio, host it on Sume (see Media inputs), keep it under 10 MB and send duration_seconds from 5 to 14.8. The lip-sync route needs a still with a clear face, or a ready avatar handle, and gives a talking clip. If you need clips longer than 14.8 seconds from one still, VEED Fabric takes up to 300 seconds at $0.1875 per second at 720p.

Worked examples around the edge

Audio of 4.9 seconds is refused, because the lower bound is 5 seconds. Audio of 5.0 seconds is accepted and costs 5 x $0.10 = $0.50 at 768p. Audio of 14.8 seconds is accepted and bills 15 seconds. Audio of 14.81 seconds is refused. The window is enforced by the schema on duration_seconds, so the check happens before any reservation.

To fit a script that runs to 15.5 seconds, speed up the voice slightly or cut a few words, then re-measure; otherwise use VEED Fabric, which accepts up to 300 seconds at $0.1875 per second at 720p, so 15.5 seconds would bill 16 x $0.1875 = $3.00.

For 1,000 clips of 14.8 seconds at 768p the batch is 1,000 x $1.50 = $1,500, so check your balance first.