10-second reshoot: face swap reserves $2.76, scripted clip $1.84

For a 10-second reshoot at the standard tier, a scripted avatar video costs 10 x $0.184 = $1.84 on Sume, while an Avatar Face Swap beta job reserves $2.76, the price for 15 seconds, as of 2026-10-09. The $0.92 difference is real but small. The decision should come from what you have: a script, or a take you want to keep.

The two routes

Face swap takes an existing public source video (roughly 4-15 seconds with usable audio) and applies a ready avatar's face to it. Avatar video takes a script and produces a new talking video of the avatar. One keeps the original performance, and the other writes a new one.

When the swap is the better answer

Pick face swap when the delivery is the value: a customer's phrasing, a founder's pacing, a gesture that sells the line. The words and the timing stay as they were shot. You also avoid rewriting a script that already works.

Pick the scripted video when the words are the value, when the source clip would break the rules (private or signed URL, no usable audio, longer than the beta window), or when you need several variants of one message. Ten script variants of 10 seconds at standard are 10 x $1.84 = $18.40, while ten swaps of ten source clips reserve 10 x $2.76 = $27.60.

Practical limits

The swap endpoint needs avatar_handle, video_url and an explicit quality (no default). It does not take prompts, transcripts, aspect ratio or duration options. The docs mark it as a Beta whose pricing may be refined before GA, so re-read the catalog before a large run.

Whoever is in the source video should know their face is being replaced and by whom. That is a consent question that no endpoint answers for you. Keep the original, the avatar handle and the output together so you can show what was changed.

Checking the source before you pay

The face-swap source has to pass the endpoint's checks. The URL must be public HTTPS and not signed or private, and the clip should be roughly 4-15 seconds with usable audio. A silent screen recording will not meet the audio expectation. Look at the clip first and trim it if needed: Sume's video trim route is $0.02 per job, small next to a $2.76 reservation.

Make the quality choice on purpose. In the beta, quality is required, so decide between standard at $2.76, plus at $3.675 and max at $8.25 before you submit.