A voice API is cheaper per minute than a white-label platform and costs more of your time. Here are the real numbers from Vapi and Retell, and when each route wins.
Lumina8 min read
If you sell AI receptionists to local businesses, building on a voice API like Vapi or Retell is cheaper per minute than a white-label platform, and it costs you far more in your own time. A voice API gives you the call engine. It does not give you a demo your prospect can open, a client portal, intake, or anyone to rebuild a prompt at 9pm when a dentist changes their hours. Which side wins depends on whether you have an engineer and more clients than you can hand-hold.
Here is the comparison with real numbers from both vendors' own pricing pages, plus where our own product is the worse choice.
What does a voice API actually cost per minute?
Both platforms charge a hosting fee and pass the model, voice and phone costs through separately.
Vapi's pricing page lists $0.05 per minute for Vapi hosting on the pay-as-you-go plan, with transcription, the language model and the voice billed by those providers on top. Their own calculator shows a 1,000 minute month landing between $82 and $129, depending on which model and voice you pick. That works out to roughly 8 to 13 cents a minute. Extra concurrent lines cost $10 each per month, and HIPAA support is a $2,000 a month add-on.
Retell's pricing page quotes $0.07 to $0.31 per minute, and its default estimator setup comes to $0.11 per minute: $0.04 for the language model, $0.055 for Retell's voice infrastructure and $0.015 for speech. Telephony is listed at $0.015 per minute. Retell includes 20 concurrent calls and charges $8 per extra line per month. Add-ons such as the knowledge base, denoising and PII removal each add half a cent to a cent per minute.
So a competent builder can run a receptionist for about a dime a minute. Remember that number, because everything else in this post is about what that dime does not cover.
What do you still have to build yourself?
The API is the engine. Everything the client and the prospect touch is yours to make.
You need a demo the prospect can try before they sign. That means a page, a phone number or a web call widget, a business profile for their trade, a prompt that handles that trade's questions, and a way to keep a simulated booking from touching a real calendar. You need a place to see whether anyone opened it. Once they sign, you need intake, a way to collect hours, services and brand assets without ten emails, and a login where the client can see status without seeing your tool stack.
Then comes the part nobody prices in. Someone has to monitor call quality, handle a transfer that drops, and change the prompt when the business changes. Neither vendor's pricing page mentions white-label terms for agencies. If you plan to put your brand on top, confirm what you are allowed to do before you build a business on it.
Lumina Sales is $49 a month for the platform. For each client build you order, you pay a wholesale setup fee and a monthly, plus voice minutes on lines that take calls, all agreed on a 30 minute pricing call before anything is built. Our top tier runs $1,150 to $1,400 setup, then $315 to $460 a month, plus $0.35 per minute on voice. You set the client price. The gap is yours.
Put that next to the API math for a client who uses 600 minutes a month. On Retell's default estimate that is about $66 in call costs. On our top tier the voice portion alone is $210 before the monthly. The API is cheaper per minute by a wide margin. That is a real gap and we will not pretend otherwise.
What the API number leaves out is the labor. If you charge $2,497 a month for a build, the margin question is not minutes. It is whether you spent 30 hours or 3 getting it live, and how many hours a month the account costs you afterward. A hundred dollar saving on minutes disappears the first time you spend an afternoon debugging a transfer. We walked through the full margin picture in what you actually keep on a white-label AI client.
When should you build on Vapi or Retell instead of buying?
Build on the API when at least two of these are true:
You have, or you are, an engineer who enjoys this work. You already have a repeatable demo and delivery process, so the platform layer would be redundant. You serve a niche with unusual requirements, such as HIPAA, a custom telephony setup or a CRM integration nobody packages. Or you expect enough volume per client that the per-minute gap matters more than your setup time.
That last one deserves a number. At $0.35 versus roughly $0.11 a minute, the difference is about 24 cents. At 1,000 minutes a month on one client that is $240 a month. If you run ten clients at that volume, you are giving up $2,400 a month in margin for convenience. At that scale, an engineer's time is probably cheaper than the premium. A small agency usually has a couple of clients and a pipeline, not ten accounts at that volume.
When is the white-label route the better call?
Use the packaged route when selling is the bottleneck, not building. The thing that wins deals is a live demo the prospect uses themselves, and a demo only helps if you can produce one in minutes for each trade you pitch. If you want the mechanics, see how long it takes to set up a white-label AI demo. On Lumina Sales you pick the trade from 100+ vertical packs, type the prospect's business name and send a link, and you can see what happens when they open it.
It is also the better call if you are a rep who does not write code, if you want a branded client portal from day one, or if you would rather spend your evenings on outbound than on prompt debugging. The live demo is on our homepage if you want to judge the quality of the product yourself before deciding anything.
Who is the white-label route not for?
Be honest with yourself on four points.
First, if your clients will run high call volume, our voice rate costs you real margin, as shown above. Second, you do not control the underlying infrastructure. If you need a specific telephony carrier, a custom model or deep control over the call flow, a raw API gives you that and a packaged platform does not. Third, the plan has caps: 100 live voice minutes, 500 chat messages and 50 website audits a month, with one published demo at a time. Those are demo limits, not client limits, but you should read them on the pricing section before you commit. Fourth, if you want to resell a product and own the stack end to end, you are building a software company, and a white-label subscription is not that.
Can you do both?
Yes, and plenty of agencies end up here. Use a packaged demo and fulfillment path to win the first clients and learn what trades actually buy. When one trade repeats and volume grows, move that single trade's delivery onto your own API build and keep the demo layer for new prospects. The mistake is doing it in the other order: building a custom stack first, then discovering you have nothing to show a buyer on a Tuesday afternoon.
FAQ
Is Vapi or Retell cheaper for an AI receptionist?
On the published numbers they land close together. Vapi's calculator shows $82 to $129 for 1,000 minutes, and Retell's default estimate is $0.11 per minute. Your real cost depends on the model and voice you choose, so run your own call length through each calculator.
Do Vapi and Retell offer white-label for agencies?
Neither pricing page we read mentions white-label terms. That does not mean it is impossible, it means you should ask their sales teams in writing before building a client-facing product on top of them.
Why does a white-label platform cost more per minute?
You are paying for the layer around the call: the demo, portal, fulfillment and onboarding. Our voice rate is $0.35 per minute on builds that take calls. That is higher than raw API cost, and it is the price of not building and running the rest yourself.
Can I start on a platform and move to my own API stack later?
Yes, though you will rebuild delivery for each client you move, so start with the highest-volume account. Check your terms with us on the 30 minute call before you plan a migration.
How many minutes will my demos use?
Our plan includes 100 live voice minutes a month with a 5 minute cap on any single call, which is about 40 demo calls. We broke that down in will 100 voice minutes cover your demo pipeline.
Next step
Take your last two prospects and estimate their monthly call minutes. Multiply by $0.11 for the API route and by $0.35 for ours, then add the hours you would spend building and maintaining each. If the hours cost more than the gap, start the $49 plan and build one demo this week. If the minutes win, sign up for Retell or Vapi and keep this post for the demo and portal checklist above.
