Per-second lip-sync and motion bands: Fabric, H3 Max, Kling on Sume
Fabric holds up to $56.25, H3 Max lip-sync up to $3.00 and Kling motion control up to $4.73. Read the catalog bands and the per-second rates behind them.
What can a per-second Sume model hold at maximum?
For per-second models, the catalog maximum is the rate times the longest input the endpoint accepts. VEED Fabric 1.0 reaches $56.25 at 300 seconds, MiniMax H3 Max lip-sync reaches $3.00 and Kling 3.0 motion control reaches $4.73.
These are reserve ceilings. A job that runs 10 seconds reserves and captures against 10 seconds, not 300.
What are the rates and bands?
Values are USD, margin included.
| Model | Rate | Min | Estimate | Max |
|---|---|---|---|---|
| VEED Fabric 1.0 | $0.1875 per second at 720p | $0.10 | $0.94 (5 s) | $56.25 (300 s) |
| Experimental avatar fabric | $0.3024 per second | $0.57 | $1.21 | $4.54 |
| MiniMax H3 Max lip-sync | $0.10 per second at 768p | $0.32 | $0.50 | $3.00 |
| Kling 3.0 motion control | $0.1575 per second | $0.16 | $0.79 | $4.73 |
What limits apply to the inputs?
H3 Max lip-sync needs audio between 5 and 14.8 seconds, and its price varies by 480p, 768p and 1080p. Kling motion control accepts up to 30 seconds. Fabric bills per whole audio second, rounded up, at 720p.
Because of the rounding, a 5.2 second clip on Fabric bills as 6 seconds, or $1.125.
How do I compute my own job?
Multiply the rate by your seconds, rounding per the model's rule. A 12-second Fabric clip is 12 x $0.1875, or $2.25. A 10-second Kling motion clip is $1.575. The estimate in the catalog is a starting point for a default request only.
Your wallet needs to cover each in-flight job's reserve at once. Read generation admission for how reservation, capture and refund work.
Does Sume cap my spend per call?
There is no per-call price cap field on these endpoints. The tools you have are the length of the input, which you control, and the wallet balance. Formats and scheduled actions have their own caps, but those are separate features.
What does a worked example look like?
A creator makes ten 8-second lip-sync clips. On H3 Max at $0.10 per second that is $0.80 each, $8.00 for the batch. On Fabric at $0.1875 per second it is $1.50 each, $15.00 for the batch, with each clip rounded up to whole seconds.
The two models take different inputs, so cost is not the only question. H3 Max needs audio from 5 to 14.8 seconds, while Fabric accepts up to 300 seconds. Choose by input first and price second.
Sume does not average or smooth these rates. Each job bills by its own seconds.
Sources
Related posts
More in Pricing
- Pika commercial license: Creator has it, Starter and Free do not
Pika's pricing page lists a commercial license on Creator and above, none on Free or Starter. What that means for client work, and what Sume's paid plans say.
- Pika credit packs: 60 credits per dollar, from $25 for 1,500
Pika sells one-time credit packs at 60 credits per dollar, from $25 for 1,500 to $1,000 for 66,000. How packs differ from a plan, and how Sume top-ups work.
- Pika Seedance 2.5 clip cost: 122 credits is $1.36 on a plan
Pika lists Seedance 2.5 at 122 credits for a 720p 5-second clip: about $1.36 at plan rates, $2.03 on credit packs. The math, and how Sume prices the same model.
- generation_admission_preview: check before paid hosted MCP calls
Sume's hosted MCP has generation_admission_preview, dry_run and max_spend_usd. What each one checks, when to use it, and when a plain single create is fine.
Written by Sume