Voice AI API pricing, per engine and per second
Wixzel Voice costs $0.0466 to $0.1851 per connected minute, depending on the voice engine, billed per second of actual usage from one prepaid balance. One API key, one balance, every voice engine; no subscription, no seats, and a $5 minimum top-up.
How much does an AI phone call cost per minute?
An AI phone call on Wixzel Voice costs $0.0466 to $0.1851 per connected minute, depending on the voice engine, plus your carrier’s rate for the SIP trunk. The figures include the flat $0.012 per minute orchestration fee and are generated from the price book the meter bills against. The meter charges per second, so a 90-second call is billed for 90 seconds, not two minutes.
| Engine | 1 min call | 3 min call | 5 min call |
|---|---|---|---|
classic | $0.0867 | $0.2601 | $0.4335 |
sarvam | $0.0623 | $0.1869 | $0.3115 |
gemini-live | $0.0466 | $0.1398 | $0.2330 |
deepgram-agent | $0.1851 | $0.5553 | $0.9255 |
Engine cost at typical speech rates, plus your carrier’s per-minute rate for the trunk, which Wixzel Voice neither includes nor marks up. A web or app session on the same engine costs the same, with no carrier cost at all.
Pick an engine, or compose one
Prices are per connected minute at typical speech rates, including the flat $0.012 orchestration fee. The table is generated from the same price book the meter bills against, so it cannot drift from what you are charged.
| Engine | Pipeline | Per minute |
|---|---|---|
classic | Deepgram + GPT-4o-mini + ElevenLabs | $0.0867 |
sarvam | Sarvam, Indian languages end to end | $0.0623 |
gemini-live | Gemini Live, native audio | $0.0466 |
deepgram-agent | Deepgram Voice Agent | $0.1851 |
Telephony is not included and not marked up. Phone calls run over your own SIP trunk, so you keep your carrier rates and your carrier relationship. Web and app sessions are billed at the same per-minute rates, with no telephony at all.
Compose your own
Name a model for each stage. The speech-to-text sets the family, the other two stages follow it, and the price is what those three add up to per minute.
Runs on the Deepgram · OpenRouter · ElevenLabs pipeline, which the speech-to-text selects. At typical speech rates, including the platform fee; metered per second, token and character, and billed for the model that actually ran. Within a family the pipeline picks the variant for the agent’s language. How composing works
{
"voice": {
"stt": {
"model": "deepgram/nova-3"
},
"llm": {
"model": "openrouter/gpt-4o-mini"
},
"tts": {
"model": "elevenlabs/eleven_turbo_v2_5"
}
}
}How many calls you can run at once
Concurrency is the number that decides whether this fits, and it is not a plan tier. Every account starts at five simultaneous calls; if you need more, that is a conversation rather than an upgrade button.
Concurrent calls per account
The default every account starts on. Enough to build, test and run a small line. Raised on request. It is a deliberate change on our side, not a plan tier.
Refused, never degraded
Past your limit, the next call is refused with a clear error and the calls already running are untouched. Nobody gets choppy audio because somebody else got busy.
Queue
There is no waiting room. A refusal is immediate and explicit, so your own retry logic decides what happens next rather than a hidden queue deciding for you.
Concurrency is not billed. You pay for connected seconds whether one line is busy or forty, so a higher limit costs nothing until it is used. The limit exists so we can size the platform honestly rather than oversell it.
Tell us what you need: how many lines busy at peak, roughly how many minutes a month, and which carrier you are on. Those three answers are usually enough for us to commit to a number in one reply.
Every charge is traceable to a second of audio
Credit is reserved before a call is placed and debited as it runs. Every micro that leaves the balance is explained by a usage row, so a bill is answerable without asking support.
Reserved at admission
Credit for the first stretch of the call is held before a port is opened or a provider socket is created. An account that cannot fund it is refused with 402, so a call that never happens costs nothing.
Debited as it runs
Speech seconds, tokens and characters are metered against the balance every few seconds while the call is live. Run low and the agent warns your caller; run out and it says goodbye rather than dropping the line.
Settled at hangup
The hold is released and only what was used stays debited, one row per component, reconciled against the call’s real duration.
$ curl .../v1/billing/balance
{
"object": "balance",
"balance_display": "$24.75",
"held_micros": 250000, # held by live calls
"available_micros": 24500000
}Amounts are integer micro-USD, so a charge of $0.000021 has a representation that agrees with the ledger. Every response carries a display string beside it.
No surprise invoices
Prepaid. You cannot be billed for more than you added.
Prices frozen per call
A call is priced at admission and finishes on those rates, so a price change never moves a meter mid-conversation.
Under-billed on our failures
If we crash mid-call you are charged for the seconds we can prove, not the amount we reserved.
Refused before it costs you
A call that cannot fund its first 15 seconds is refused before anything is allocated. At about a minute of runway the agent warns your caller and then hangs up cleanly.
Or run it on your own server
With a self-host license you install Wixzel Voice on servers you control and pay your voice providers directly, with no orchestration fee on calls. It costs $499, paid once, and every future release is included.
Pricing questions
- What does Wixzel Voice cost per minute?
- From $0.0466 (gemini-live) to $0.1851 (deepgram-agent) per connected minute at typical speech rates, including the flat $0.012 per minute orchestration fee. Every figure is generated from the price book the meter bills against and rounded up, never down. Telephony runs over your own SIP trunk at your carrier's rate and is not included or marked up.
- Is there a subscription, a seat price or a minimum monthly spend?
- No. Credit is prepaid from a $5 minimum top-up and spent per second of actual usage. An account with no calls spends nothing.
- How are web and app sessions billed?
- Exactly like a phone call on the same engine: the same per-minute price, from the same balance, per second. A session counts toward your concurrent-call limit, appears in your call log with
channel: "web"and ends with the samecallCompletedwebhook. There is no telephony cost, because there is no phone line. - How am I billed?
- Prepaid. You add credit, calls spend it per second of speech, per token and per character, and you cannot be billed for more than you put in. The minimum top-up is $5.
- What happens when credit runs out?
- A call that cannot fund its first 15 seconds is refused with a 402 before anything is allocated. Mid-call, at about a minute of runway, the agent warns your caller, then speaks a closing line and hangs up. It does not drop the line mid-sentence.
- Can I retry a request safely?
- Send an
Idempotency-Key. It is required onPOST /v1/calls,POST /v1/agents/{id}/test-callandPOST /v1/billing/topups, and a replay with the same key and body returns the original response instead of dialling twice. Keys are kept for 24 hours.POST /v1/campaigns/{id}/starttakes no key, and starting a stopped campaign dials every lead on it again. - Is there a free trial?
- There is no time-limited trial and no free tier. Credit is prepaid from $5, a short test call costs a few cents, and the Credits and Refunds policy explains what can be refunded. You can talk to each engine from the landing page before adding any credit.
- How many calls can run at once?
- Five per account by default, phone calls and web sessions counted together. Past the limit the next call is refused with a clear error and the calls already running are untouched. The cap is raised on request; email the address on the contact page with the volume you expect.
- Is there a way to pay once instead of per minute?
- Yes. The self-host license is $499, paid once, to run Wixzel Voice on your own servers with your own voice provider accounts and no orchestration fee on calls. The installer, every future release and the source code are included. You then pay your providers and your carrier directly.
Start with a curl
Sign up and create a key. Point a SIP trunk at us for phone calls, or mint a realtime session to put the agent in your app.