Why is my real voice AI cost per minute higher than the advertised rate?

Because the advertised rate is almost always a single line item, not the whole bill. Most voice AI platforms and wrappers quote you their platform markup, then pass through telephony, language model inference, and text-to-speech plus speech-to-text as separate invoices you assemble yourself. On top of that sits developer time to wire the pieces together and keep them running. When you stack all of it, the true blended cost per minute is commonly two to three times the number you were quoted. This calculator adds every layer, divides fixed monthly costs across your call minutes, and shows the real per-minute figure next to the advertised one so the gap is impossible to miss. The point is not to scare you, it is to price your own service with the real number instead of the marketing number.

What are the five invoices in a typical voice AI stack?

First, the platform or wrapper fee, which is the per-minute rate most providers advertise. Second, telephony, usually Twilio, billed per minute for the actual phone call. Third, the language model, billed per token and effectively per minute of conversation. Fourth, text-to-speech and speech-to-text, the voice and transcription layer, often a separate vendor with its own per-minute rate. Fifth, developer time, the human hours to integrate those four services, handle failures, and maintain the glue as APIs change. Four of those are literal invoices from four vendors and the fifth is payroll or contractor time. Reconciling them every month is its own tax on your margin. This tool counts the cost of each so you can see the total you are actually carrying, not the one slice a sales page shows you.

How does this calculator compute the blended cost per minute?

It sums the four per-minute cost layers, the platform rate, telephony, the language model, and text-to-speech plus speech-to-text, into a variable cost that scales with every minute of talk time. Then it takes your fixed monthly costs, the platform subscription and developer hours at your hourly rate, and divides them across your monthly call minutes to get a per-minute fixed component. Add the two together and you have the true blended cost per minute. It also multiplies the gap between that blended figure and the advertised rate by your minutes and by twelve to show annual overpay, the money leaking out because you priced against the wrong number. Every field has a sensible default so you can get a real answer in seconds, then tune it to your exact stack and volume for a precise figure.

How much do developer hours really add to cost per minute?

More than most operators expect, especially at lower call volume. Developer time is a fixed monthly cost, so it hits your per-minute number hardest when minutes are low. Eight hours a month at $75 an hour is $600, and spread across 2,000 minutes that is $0.30 added to every single minute before you have paid for a single API call. At 500 minutes the same $600 adds $1.20 per minute. This is why a duct-taped stack quietly destroys margin for solo operators and small teams, the dev tax does not shrink until volume is large. Lower the minutes in this calculator and watch the blended cost climb, then raise them and watch it fall. It is the clearest way to see why a platform that removes the developer dependency changes your unit economics, not just your workload.

How does Hermes pricing compare to a five-invoice stack?

Hermes bills one number and sends one invoice. Voice usage is $0.24 per minute all in, covering telephony, the model, and the voice layer, with no separate vendor bills to reconcile and no developer required to keep it running. Plans start at $149 a month for Starter with 300 included minutes, $399 for Business with 1,000, and $699 for Agency with 2,000, each covering your whole client base rather than a single agent. The comparison in this tool is kept honest, it shows your true DIY monthly total, including developer time, next to the Hermes monthly total for the same minutes, and tells you which is lower. Even when the raw per-minute figures are close, one invoice and zero developer hours change the real cost of running the business, and white-label is included rather than another project to build.

What should I do with the true cost number once I have it?

Price your service off it, not off the advertised rate, so your margin survives contact with reality. If your true blended cost is $0.30 a minute and you sell a client 1,000 minutes a month, that floor is $300 in cost before your time, and your retainer has to clear it with room to spare. Use the annual overpay figure to decide whether the current stack is worth maintaining or whether consolidating onto one platform pays for itself. Then book a 20-minute walkthrough and we will map your real per-minute cost to a Hermes plan, size the included minutes against your call volume, and show you what your margin looks like with one invoice instead of five. Want the full breakdown for your agency? Book 20 minutes and we will build it with you.

free tool

What does a voice minute really cost?

The rate you were quoted is one invoice of five. Add telephony, the model, the voice layer, and developer time, and see your true blended cost per minute, your annual overpay, and how it compares to one flat number.

/ advertised platform rate ($/min)The per-minute rate the platform or wrapper quotes you. This is the number people plan around./ telephony ($/min)Twilio or your carrier, billed per minute for the actual call. Around $0.014 is typical./ language model ($/min)Model inference, billed per token, converted to a per-minute rate. Varies widely by model./ text-to-speech + speech-to-text ($/min)The voice and transcription layer, often a separate vendor with its own invoice./ platform subscription ($/mo)Any flat monthly seat or platform fee on top of usage. Zero if usage-only./ developer hours per monthHours spent integrating and maintaining the glue between these services each month./ developer hourly rate ($)Your cost per developer hour, contractor or your own time valued honestly./ call minutes per monthTotal talk minutes across your book. Fixed costs get spread across this.

/ your true cost per minute

True blended cost per minute

$0.554

You were quoted $0.130. Real cost is 4.3x that: $0.254 variable + $0.300 fixed per minute.

Annual overpay vs the advertised rate

$10,176

The gap between the quoted rate and reality, across 2,000 minutes a month for a year.

Bills you reconcile every month

5 cost sources → Hermes: 1 invoice

Fixed costs run $600 a month before a single API call. On Hermes that is zero developer hours.

Same minutes on Hermes (Agency)

$699/mo · $0.349/min

Your true DIY total is $1,108/mo, so Hermes is $409/mo cheaper here, and it is one invoice with zero dev hours.

Book a 20-min walkthrough to cut this number, and we will map your true per-minute cost to a plan with your minutes sized to your call volume.

Book a 20-min walkthrough

how it works

The advertised rate hides four more bills.

When a voice AI platform quotes you a per-minute rate, that number is almost never the cost of running a minute of conversation. It is the platform markup, one line on a bill that has four more. The phone call itself is telephony, usually Twilio, billed per minute. The conversation runs on a language model billed per token, which works out to a per-minute rate of its own. The voice you hear and the words the agent understands come from a text-to-speech and speech-to-text layer, often a separate vendor with a separate invoice. And gluing those services together, keeping them running, and fixing them when an API changes is developer time, a fixed monthly cost whether you call it payroll or a contractor.

This calculator stacks all of it. The four per-minute layers add up into a variable cost that scales with every minute of talk time. The fixed monthly costs, the platform subscription and the developer hours, get divided across your call minutes to produce a per-minute fixed component. Add the two and you have your true blended cost per minute, shown right next to the advertised rate so the gap is plain. The tool then multiplies that gap by your minutes and by twelve, because the real damage of pricing against the wrong number is not one call, it is a year of them. Developer time is the quiet killer here: it is fixed, so at low volume it can add a dollar or more to every minute before a single API call is paid for.

Hermes exists to collapse those five invoices into one. Voice usage is $0.24 per minute all in, telephony, model, and voice layer included, with no vendor bills to reconcile and no developer to keep the stack alive. Plans start at $149 for Starter with 300 minutes, $399 for Business with 1,000, and $699 for Agency with 2,000, each covering your whole client base under your own brand. The comparison below is kept honest: it shows your true DIY monthly total against Hermes for the same minutes and tells you which is lower.

frequently asked

Common questions.

next step

Run your agency on Hermes.

One platform. Your brand. From $149/month. First agent live in 72 hours.

Book a 20-min walkthrough See pricing

What does a voice minute really cost?

The advertised rate hides four more bills.

Common questions.

Run your agency on Hermes.

/ Keep reading