Three 15 s avatar hooks at standard ($8.28) vs one at max ($8.25)

On Sume Avatar 1.0, three 15-second hook variants at the standard tier cost 3 x 15 x $0.184 = $8.28, which is within three cents of a single 15-second render at the max tier (15 x $0.55 = $8.25).

The comparison

The catalog prices standard at $0.184, plus at $0.245 and max at $0.55 per second, without a product image. The docs describe standard as the fastest execution path and max as the highest quality, slower turnaround.

The numbers hold only for a script of 15 seconds. If your hook is closer to 8 seconds, three standard hooks cost 3 x 8 x $0.184 = $4.416, and one max hook costs 8 x $0.55 = $4.40. The ratio between the tiers is what stays fixed: max is about 3 times the standard rate ($0.55 / $0.184 = 2.99).

Run the variants, then upgrade the winner

The avatar video previews route lets you approve first-frame stills and then call generate-video with a quality override. That override affects only the final render tier, and the stills are tier-independent. A sensible loop is: draft the three hooks at standard, pick the winner by watching, then render the winner again at plus or max. That costs $8.28 + $3.675 = $11.955 for a plus finish of a 15-second hook, or $8.28 + $8.25 = $16.53 for max.

Avatar video has a 4-second minimum window, so very short hooks need a line long enough to fill 4 seconds or a silence beat. Multi-scene video_inputs let you build a hook, a demo and a call to action in one render, with each scene's duration set on the scene.

• Hooks that differ in the first line need separate jobs.
• Keep the avatar_handle and the scene the same so only the wording changes.
• Change the script, not the tier, between tests.

When max is the right call

If you already know the script and the face and only need the best output for a paid placement, a single max render skips the drafts: $8.25 for 15 seconds. The three-hook approach is for when the script is the unknown.

Write the three hooks so that they differ in one way only: the opening line, the claim, or the offer. If all three change, you cannot tell which change won. Keep the avatar, scene and aspect ratio fixed, use a distinct Idempotency-Key per hook, and label the results with the hook id before you look at any performance data. After the test, rerender only the winner at the tier you want to publish. The loser drafts cost $5.52 together at standard, which is the whole price of the lesson.