Prototype01
Free balance
$5
on signup
Enough session time to connect a real prototype before adding a payment method.
- $5 starting balance
- All public models
- 4 concurrent session
WFPricing / 03
Pay only while the world is alive.
Billing begins at the first generation_started event. Queueing and startup are free. After generation starts, paused time is billed until release or session end.
Connectingfree
Queuedfree
Before generationfree
Generation startedmetered by second
Endedmeter stopped
01 · Live meter
One runtime.
One honest meter.
Choose a model, billable time, and concurrency. The estimate moves on the same unit as the API: session seconds.
Windflow runtime estimatorMeter online
Select runtime
Estimated burn
Helios
Priced in USDReady state only
$0.102
1 × 1 min billable time · 60 session seconds.
- Per session
- $0.102
- USD / hour
- $6.12
- Free runs
- 49
1 min
01
01 session50 sessions
No generation-count abstractionIndicative private beta rate
02 · Access
Start free. Scale without changing the meter.
Every access mode uses the same billable-time billing model. Capacity changes; the unit does not.
Production02
Pay as you go
$0.0017
per session-second, from
Pay by the second from the first generation_started event until release or session end. No subscription or seats.
- Ready-time billing
- Recoverable sessions
- Concurrency on request
Design partner03
Reserved scale
Custom
reserved capacity
For teams shipping interactive video with dedicated capacity and private deployments.
- Reserved GPUs
- Private runtimes
- Direct engineering access
Indicative private-beta rates · final GA pricing may shift ±15%Meter starts · first generation
03 · Billing FAQ
What changes a balance. What does not.
These answers describe the session response's billing contract. Rates remain discoverable at the API instead of being hidden in plan fine print.
Q01When does the meter actually run?
Billing starts at the session's first generation_started event and stops when the session is released or ends. Queueing, worker startup, time before generation starts and the documented reconnect grace are not billed. After generation starts, paused or unwatched time remains billable; release the owner session to stop the meter.
Q02Does pausing video stop the meter?
No. pauseTrack only stops delivery to that connection, and a generation pause can still leave the GPU reserved. Release the owner session when the experience is finished or idle.
Q03Why can I get a 402 while my balance is positive?
Session creation reserves the model rate multiplied by its maximum session ceiling. Concurrent reservations reduce the spendable balance. When a session ends, actual billable seconds settle and the unused reservation returns.
Q04Do additional viewers cost more?
No. One session is metered once regardless of how many authorized guests receive it. Separate concurrent sessions each carry their own meter and reservation.
Q05Why do models have different rates?
Rates follow the GPU and runtime profile needed to keep each model interactive, not an output-count abstraction. Every rate is published in US dollars per billable session-second; current public rates are also exposed by GET /v1/pricing.
Q06What happens to charges if a runtime fails?
Queueing, startup and time before generation starts are free. If a session fails, it ends and settles against the billable seconds already recorded; the unused reservation returns to the account balance.
Q07Are uploads or recordings separate line items?
During private beta, the published public rate is the billable-session meter. Temporary recordings expire after 24 hours; high-volume exports, dedicated storage, or private delivery are scoped with Reserved scale.
Q08Can I test a real session before paying?
Yes. Approved accounts receive $5 on signup, usable across public models. The estimator above shows how many runs fit inside that grant.