Appearance
Billing
xVoice is pay as you go. You buy credit in advance, and each request is deducted from your balance as it is used. There are no subscriptions or minimum commitments.
Prices are on xvoice.dev/#pricing. For volume pricing or a Business Associate Agreement, contact sales.
How usage is metered
| Product | Metered per |
|---|---|
| Text to speech | Character of input, counted before generation starts |
| Speech to text | Second of audio, measured from the file |
| Realtime transcription | Second of audio sent over the session |
Usage is metered exactly, never rounded up to the minute or the request. Realtime transcription is priced separately from file transcription.
Credits
- Free starting credit. Verifying your email adds $5 of credit, which expires after 30 days.
- Top-ups are $5 to $1,000, paid by card from Billing in the dashboard. Only organization owners can buy credit.
- Purchased credit never expires.
- Spending order. Credit that expires soonest is used first, so your free credit is spent before credit you paid for.
When the balance reaches zero, requests fail with 402 insufficient_credits. Requests already in progress are allowed to finish. Anything they cost beyond your balance is taken from your next top-up.
Auto-recharge
To keep a service running without watching the balance, turn on auto-recharge under Billing. Save a card, then set a threshold and a target balance. Whenever the balance falls below the threshold, the card is charged to bring it back up to the target. The target must be at least $5 above the threshold.
The dashboard shows a banner when your balance is low or empty. There are no email alerts yet, so use auto-recharge for any service that runs unattended.
Spending limit
Each organization has a monthly spending limit, $100 by default, which you can change under Billing. When this month's spending reaches it, requests fail with 402 spending_limit_reached, even if credit remains.
It protects you from runaway usage, such as a retry loop spending real money. Set it to what a bad month should cost, not to what you expect to spend.
What is billed
If the model did the work, the request is billed.
| Situation | Billed? |
|---|---|
| A successful request | Yes |
| Rejected before processing: bad key, bad model, over a limit, file too large | No |
| The model was unavailable or at capacity, and produced nothing | No |
| You disconnected while audio or text was streaming | Yes |
| Realtime audio you sent and then cleared | Yes |
| Realtime audio sent before you disconnected | Yes |
| A realtime utterance that failed with no text sent | No |
Disconnecting does not undo work already done. Realtime sessions are billed for the audio you send, not for the transcripts you receive, because a client that disconnects never receives the final transcript.
Usage and request logs
The dashboard shows:
- Usage, totalled by day, product, model and project.
- Request logs, one row per request, with model, status, latency, cost and request id.
Both are built from the same records your balance is charged from, so they always match. The request id in a log row is the X-Request-Id header and the request_id in an error, so any request your code logged can be found in the dashboard.