17-second avatar video: $3.13 standard, $4.17 plus, $9.35 max

A 17-second avatar talking video costs $3.128 at standard, $4.165 at plus and $9.35 at max on Sume (as of 2026-10-09). Plus is the default if you leave quality out. Sume bills per second of finished video, and the docs accept scripts that Sume estimates between 4 and 60 seconds.

The cost table

The same request body works at every tier; only the quality value changes.

Spend less on the way to the final

Avatar video previews generate first-frame stills before the full render. If you approve a preview and then call generate-video, you can change quality at that step without a new preview, because the stills are tier-independent. That makes a two-step plan practical: preview, then standard to check timing, then max for the version you publish. Paying for max on a script you will rewrite costs 17 x 0.55 = $9.35 each time; standard for the draft costs $3.128.

A change to the script, scene, avatar or aspect ratio does need a new preview. Quality alone does not.

Where the money goes in a three-render workflow

Imagine a draft at standard, a revision at standard and a final at max. That is 3.128 + 3.128 + 9.35 = $15.606. Doing every render at max would cost 3 x 9.35 = $28.05, and every render at plus 3 x 4.165 = $12.495. The mix with the draft at standard saves $12.444 compared with all-max.

The preview route changes the picture again. A preview shows the first frame for you to approve, and a later generate-video call reuses it. If the first frame is what you were worried about, you may not need the draft render at all. Reserve renders for judging motion and timing, and use previews to judge framing.

Duration planning

The duration is estimated from the script, so 17 seconds is a planning number, not a promise. The result carries the final video. If you want to hit a hard length, count on the estimator and leave slack under 60 seconds. For a script that would run past 60 seconds, split it into two jobs; at plus, two 30-second jobs cost 60 x 0.245 = $14.70, the same as one 60-second video.

Keep the quality value in your job log. When someone asks why two clips from the same script look different, the tier is the first thing to check, and a log line with handle, tier and job id answers it quickly.

• Aspect ratios: 1:1, 3:4, 9:16 (default), 4:3, 16:9.
• Resolution is 720p today.
• Optional inline captions burn into the final MP4 and do not create a separate billed caption job.