Short clip lipsync: Kling bills 5 s steps, LatentSync 40 s

On Sume, a 6-second clip on Kling Lipsync is billed as 10 seconds, and a 12-second clip on LatentSync is billed as 40. Both rows set a billing rule on short clips, and neither shows up unless you read the row's note: Kling rounds the source clip up to the next 5 s, LatentSync bills at least 40 s.

Both were added to the Video Router on 2026-10-13 as video-to-video lipsync rows. Sume shows credits, which scale with the billed seconds below.

Billed seconds by clip length

The Sume repo's pricing code computes billed seconds as follows: for Kling Lipsync, the source clip length rounded up to a whole multiple of 5; for LatentSync, the clip length with a floor of 40.

What this means for a batch

Kling Lipsync is cheapest when clips land on 5 or 10 seconds exactly. Trimming a 6 s clip to 5 s cuts a full step of billed time. Because its duration is the source clip's length, not the output's, the audio length does not change the bill.

LatentSync is the opposite: it rewards long clips. Ten 4-second lines sent as ten jobs bill 400 seconds; stitched into one 40-second clip and sent once, they bill 40. If the lines have to stay separate files, a row without a floor (Sync Lipsync 3 or VEED Lipsync, both per output second) fits better.

Rules like these are why the row notes in the catalog matter more than the headline per-second figure. Two rows with the same quality can differ by a factor of eight on a four-second line, and the gap closes as clips get longer. When you plan a batch, tabulate billed seconds per clip, not clip seconds, and sum that column.

Which row to pick

• Clips of 2-10 s that you can cut to 5 or 10 s: Kling Lipsync.
• One long clip, 40 s or more: LatentSync bills it at its true length.
• Many short clips: avoid the 40 s floor; use a row billed per output second.
• Anything that needs a scene prompt: InfiniteTalk Video, 2-9 s, prompt required.

Check before you submit

Send duration as whole seconds and the row's own limit applies first: Kling accepts 2-10 s, so a 12 s clip is refused before any billing. See what duration to send when clip and audio differ and the full list of video rows. Credits are explained in Sume credits.