H3 Max Lip Sync on an 11.2 s line: $0.75, $1.20 or $2.40

An 11.2-second voice line sent to Sume's MiniMax H3 Max Lip Sync is billed as 12 seconds: 12 x $0.0625 = $0.75 at 480p, 12 x $0.10 = $1.20 at 768p and 12 x $0.20 = $2.40 at 1080p. The route bills ceil(audio seconds) at the resolution you pick, and the output length follows the audio.

The three prices for the same line

The catalog row for minimax/h3-max/lip-sync lists $0.0625, $0.10 and $0.20 per audio second at 480p, 768p and 1080p. Those include Sume's standard margin over the provider list. The only thing that differs between the three runs is the resolution field, so the cost ratio is fixed: 1080p costs 3.2 times 480p ($0.20 / $0.0625).

Prices are linear inside a resolution, so you can scale the table. A 7-second line costs 7 x $0.10 = $0.70 at 768p, and a 14.8-second line bills 15 seconds, which is 15 x $0.10 = $1.50 at 768p. The 15 billed seconds is above the 14.8-second window only because of rounding; the window itself is measured on the audio you send.

The audio window that decides eligibility

The same catalog row states the audio must be 5 to 14.8 seconds, Sume-hosted and under 10 MB. It adds that the provider rejects shorter audio and silently clips longer audio. An 11.2-second line is inside the window with room to spare; a 15-second read would lose its last 0.2 seconds without an error, so measure first.

Because the body matches VEED Fabric 1.0 (still plus audio), you can send the same still and audio to either route. Fabric costs more per second at 720p ($0.1875) than H3 Max at 768p ($0.10).

Because output length follows the audio, there is no separate duration to set on the request body for the clip. What you choose is the audio and the resolution. Keep the voice file small and clean: if the file is over 10 MB or not hosted on Sume, the job does not start.

• Window: 5 to 14.8 s of audio.
• File: Sume-hosted, under 10 MB.
• Billing: ceil of the audio length, so 11.0 s and 11.2 s differ by one billed second.

Picking a resolution

Use 480p for review drafts and for lines you will place small in a timeline. Use 768p when the face fills a vertical frame. The 1080p tier is worth it only when the clip will be shown full-screen; at $2.40 for 12 seconds it is double the 768p price.

A practical flow is to run the line once at 480p for $0.75, check the mouth timing against the audio, and rerun at 768p only if the first take is right. That costs $0.75 + $1.20 = $1.95 in total when a rerun is needed, and $0.75 when the draft is already good enough to use.

Before you submit, run three checks on the audio file: the length in seconds, the file size in megabytes, and whether the file is hosted on Sume. The length decides eligibility and cost, because 11.2 seconds bills as 12 and a 15.0-second file is clipped. The size decides whether the job is accepted. The hosting decides whether the route can read it at all. Doing the checks in code before the call is cheaper than finding out from a failed job, and it gives you the billed-seconds figure to log next to the job id.