Every MiniMax price on Sume: H3, H3 Max, lip sync, Recast

Sume sells four MiniMax-based routes: minimax-h3, minimax-h3-max, minimax/h3-max/lip-sync and h3-max-recast. Each bills the provider list price times 1.25; per second that runs from $0.0625 for 480p generation to $0.5625 for 1080p Recast.

This is the whole price sheet in one place, with 10-second totals and the rule that changes each route's bill.

The table

List prices are the fal figures recorded in the Sume docs and pricing source (dates in the notes below). The billed column is list x 1.25. Totals for 10 seconds are rounded up to the cent.

What changes the bill

For generation, duration is the lever: both rows take 5 to 15 seconds. For minimax-h3, a reference-to-video request adds $0.08 list, or $0.10 billed, for each reference image after the fifth. For minimax-h3-max the reference-to-video job sells at the per-second list.

For lip sync, the audio length sets the output and the bill is ceil(duration_seconds) times the per-second price. Audio must be 5 to 14.8 seconds, so the range is $0.50 to $1.50 at 768p and $1.00 to $3.00 at 1080p per clip. For Recast, the source length rounded up sets the bill, between 5 and 30 seconds, so 768p runs from $1.88 to $11.25 and 1080p from $2.82 to $16.88.

Where the list prices come from

The Video Router notes cite fal pages dated 2026-08-25 for H3, 2026-09-08 for H3 Max, and 2026-10-01 for Recast. The lip sync rate card is dated 2026-09-22. The docs also note that fal ran promotional prices on H3 Max, and that Sume bills the ongoing list and not the promo. Prices can change, so read the catalog row for the live number.

Every submit response carries usage.billable_amount_usd_micros, the quote for that job. Admission preview, POST /v1/generation/admission-preview, accepts the lip-sync public id if you want a quote before you submit.

• Cheapest MiniMax second: $0.0625 at 480p generation or lip sync.
• Most expensive: $0.5625 at 1080p Recast.
• 1080p generation and 1080p lip sync both bill $0.20 per second.

Choosing from the sheet

If you have no footage and no audio, start with generation. If you have a voice line and a still, lip sync costs the same per second as H3 Max at the same tier. If you have footage and want a different person in it, Recast is the only route here that keeps your source scene, and it costs 3.75 times the H3 Max 768p generation rate.

A worked bundle

A campaign for one product might use three routes. Ten 10-second H3 Max takes at 768p are $10.00. Two 12-second lip sync clips of a spokesperson still at 768p are 2 x $1.20 = $2.40. One 15-second Recast of the hero ad at 768p, for a second market's creator, is $5.63. The total is $18.03 before reruns.

Move the same bundle to 1080p and it is $20.00 for the takes, $4.80 for the lip sync and $8.44 for the Recast, a total of $33.24. The sheet makes the cost of each step visible so you can decide which one deserves the higher tier.

Reading the table

All figures are Sume prices, which are the provider list price times 1.25, rounded up to the cent on the total. The submit response carries usage.billable_amount_usd_micros, so you can confirm the amount for a given request before the job runs, and a failed job is refunded.

If a row you expect is missing from your account, read the live catalog at GET /v1/video-router/models rather than relying on a blog table, because rows depend on which providers are configured.