New avatar plus an 11-second face swap: $2.97 standard, $7.00 max

Starting from nothing, an avatar plus an 11-second face swap costs $2.974 at standard, $3.645 at plus and $7.00 at max on Sume (as of 2026-10-09). The avatar is $0.95 once. The swap is 11 seconds times $0.184, $0.245 or $0.55.

The two jobs, in order

Face swap needs a ready avatar handle, so the first job creates the avatar and the second job uses it. Both are asynchronous jobs: submit, poll status, read the result. Wait for the avatar job to complete before you submit the swap.

What the avatar step needs

Avatar creation takes a top-level avatar_handle and an input union. The three input types are a text prompt, structured props (ethnicity, sex, age) and a photo with a public HTTPS image URL. Sume strips a leading @ from the handle and stores it without one. The prices are the same for all three types.

The handle is your key for every later job, so choose it before you start. If you plan to swap one avatar onto a batch of clips, the $0.95 is paid once. For five clips of 11 seconds at standard that is 0.95 + 5 x 2.024 = $11.07, which is $2.214 per clip all-in.

If the swap needs a redo

A redo costs the swap again but not the avatar. A second 11-second swap at standard is another 11 x 0.184 = $2.024, so a clip that takes two attempts is 0.95 + 2 x 2.024 = $4.998. Because the avatar already exists, nothing in the second attempt creates a second $0.95 charge.

If the first result has a problem with the face itself, the issue may be the avatar, not the swap. In that case a second avatar built from a different photo or prompt is a $0.95 experiment. Compare both on the same source clip.

Source clip rules to check first

The 11-second clip must be a public HTTPS URL, must carry usable audio, and must sit inside the roughly 4 to 15 second beta window. The swap does not take a script. If you want a new script spoken by the avatar, use Generate avatar video instead.

If the same avatar will be used for talking videos too, the $0.95 is shared between the uses, so the swap's share of the avatar cost falls with each extra job.

As a budget line, write the two jobs separately: avatar creation $0.95 and the swap at 11 x the tier rate. That keeps the one-time cost visible when you later spread it over many clips.

Keep the handle consistent between the two jobs. A typo in avatar_handle on the swap request fails the job instead of using the avatar you just paid for.