Limits and Credits

Limits vary by plan. GET /v1/billing/plans returns the published plan limits. GET /v1/billing/quota returns your current credit balance and whether the account can start a realtime session.

​

Realtime sessions

Realtime plans set the maximum number of concurrent sessions and the maximum length of each session. Exceeding the concurrency cap returns 429 quota_exceeded with error.code: "concurrent_sessions".

​

Other session limits

The idempotency window applies to supported POST endpoints. Reuse a key only when the request is unchanged. Run and session retries replay the original response. If you send max_duration_seconds, Protoface applies the lower of your request and your plan limit. For early capacity planning, divide daily active users by 250 to estimate concurrent sessions. Protoface may also limit how quickly new sessions start during a ramp-up; active sessions are not affected.

​

Realtime credits

Usage is rounded up to the nearest minute. One credit equals one rounded-up session minute. If the account has no usable credits, session creation returns 402 quota_exceeded with error.code: "insufficient_credit_balance". Free monthly credits do not roll over. Purchased credits roll over without expiry.

​

Model inference

When you submit a run, Protoface reserves the maximum_reserved_credits shown in the estimate. It charges those credits when the run completes and releases them if the run fails. Each plan also limits concurrent inference runs that produce video. If the account is at its limit, the API returns 429 quota_exceeded with error.code: "video.concurrency_exceeded". Wait for an active run to finish before submitting another one. Canceling a queued run releases its reservation. Canceling after work starts charges the reserved amount.

​

Metadata keys

metadata is customer-owned JSON. Session keys beginning with _avatar. are managed by Protoface. Model-run keys beginning with protoface_ are reserved and rejected.

​

Next