One hundred live voice minutes is about forty demo conversations, or twenty if every call runs the full five minutes. Here is how the meter works, how to plan a month of demos against it, and why the single published demo limit matters more than the minutes.
Lumina8 min read
One hundred live voice minutes a month covers about forty demo conversations, because a prospect who pokes at a demo receptionist spends two or three minutes on it, not ten. If every call ran to the five minute ceiling you would get twenty. Either number is more than most people closing one to three deals a month will use. The limit that shapes your week is not the minutes. It is the single published demo.
That is the honest version. Below is how the meter works, how to plan a month of demos against it, what your prospect sees if you run out, and the point where this plan stops fitting.
How does a voice minute get spent?
A minute comes off your allowance while someone is in a live conversation with the voice agent on your demo page. That is the whole rule. Browsing the page costs nothing. Reading the trade copy costs nothing. The live demo starts metering when a visitor clicks into a call and starts talking.
The other two meters are separate and never touch your voice budget. Chat runs on its own count, 500 messages a month, and one message is one message. Website audits run on theirs, 50 a month, and one audit of one domain is one audit. So a prospect who opens your link, audits their own site, chats with the agent for a while, then calls it, has spent one audit, a handful of messages, and only the call portion of your voice minutes.
There is a hard five minute ceiling on any single call. That matters more than it sounds. It is not there to ration you, it is there so a prospect who wanders off with the tab open cannot drain a third of your month in one sitting. The full limit table sits next to the plan rather than buried in a terms page.
So how many demos is 100 minutes, really?
Depends on call length, so here is the arithmetic in both directions.
At the five minute ceiling on every call, 100 minutes is 20 calls. At two and a half minutes, which is closer to what people do when they are testing whether an AI receptionist sounds right, it is 40. At ninety seconds, which is what a skeptical owner spends before he decides it is real, it is about 65.
Now map that against a pipeline. Say you send ten demo links a month. Six of those prospects open the page, four talk to the agent, and two of the four forward the link to a partner or an office manager who tries it as well. That is six voice conversations from ten sent links. Call it fifteen minutes. You would have to run that volume six times over before minutes became your constraint.
The people who hit the voice cap usually do not have big pipelines. They are the ones using the demo as their own sandbox, calling it fifteen times in an afternoon while they tune the greeting and the price bands. If that is you, tune in chat mode where the meter is 500 messages, then confirm the voice once.
What happens when you hit the cap?
The live nodes pause and everything else on the page keeps working. Your prospect sees a neutral line saying the demo is not accepting calls right now. They do not see a billing notice, they do not see your plan name, and they do not see the word Lumina. The client portal, the audit, the rest of the page, all of it stays up.
What does not arrive is an invoice. The plan never meters you past the cap. That is a different arrangement from how most of this market bills, and it is worth understanding before you compare price tags.
Assemble your own stack and you pay per minute for as long as you own the account. Vapi charges a flat 5 cents a minute for its own orchestration and passes transcription, model, and voice costs through at cost on top, with phone numbers at $2 a month. Retell advertises $0.07 to $0.31 a minute depending on which model you put behind it, with its own voice infrastructure line at $0.055 a minute before the LLM and the telephony. ElevenLabs bundles minutes into monthly tiers, 275 minutes on the $22 Creator plan and 1,238 on the $99 Pro plan, then charges $0.080 a minute for overage and $0.160 a minute for burst above that. OpenAI publishes no per-minute number for realtime voice at all. It bills GPT-Realtime-2.1 audio by the token, $32 per million in and $64 per million out, which makes the cost of a call something you model rather than something you read off a page.
Every one of those is a fine way to sell voice. None of them has a stop. On a metered account, a demo link forwarded around a twelve person company on a slow Thursday is a line item. On this plan it is a paused node. We went a layer deeper on what voice actually costs per minute if you want the underneath of that.
The limit that actually bites: one published demo at a time
Here is the constraint people trip over, and it has nothing to do with minutes.
You get unlimited drafts and one published demo. The link you send is one link, pointed at one prospect's demo. Publish the next one and the previous link stops serving.
Two ways to work with that. Run a one at a time cadence: publish for the call, leave it up through the follow up window, swap it when the next prospect is ready. Or build several drafts on Monday and publish each the morning of its call. Both work. Neither is free.
Where it hurts is parallel outbound. Five live links out at once, each sitting in a different inbox for a week, is not what this plan does. If that is how you sell, better to know now than after you have sent the fourth link and broken the first three.
Watch your follow up window too. A prospect who opens your link nine days later opens whatever is published then, which might be somebody else's business name. The fix is boring. Know when you are swapping, and get the demo sent while the conversation is warm.
Client builds are metered separately, and that is the number to watch
Do not confuse the plan allowance with what a delivered client system costs to run. Different lines entirely.
The 100 minutes are for demos, on our infrastructure, with no phone number to buy and no API key to obtain. A build you order for a real client, where a real business line takes real calls, carries a wholesale setup fee and a monthly, plus per minute voice where the build answers the phone. On the top tier that voice rate is $0.35 a minute, agreed with you on a thirty minute call before anything gets built.
That is the number that scales with your client's business, and the one to price against. The demo allowance is not. Our margin math post works through what is left after it.
Who should not buy this plan
If you need more than one live demo link at a time, this is the wrong tool and no quantity of minutes fixes it.
If you run high volume outbound, hundreds of touches a week with a demo attached to each, you will hit the published demo limit first and the minutes eventually. You would be better off building your own metered stack where the ceiling is your credit card.
And if the plan is to leave a public demo up on your own website as permanent lead capture, 100 minutes is not that budget. Strangers on the internet are not disciplined callers. That use case wants a metered account.
FAQ
Do unused minutes roll over?
No. The allowance resets each cycle and nothing carries forward. Treat 100 as a monthly ceiling, not a bank.
What does my prospect see if I run out mid conversation?
The call ends and the node shows the neutral not accepting calls message. It names no plan, no cap, and not us. It reads like a page between states, because that is all it is. We covered everything a prospect can and cannot see separately.
Can I buy more minutes?
The plan sells no metered overage, on purpose, so that no forwarded link can produce a bill you did not expect. If you are pressed against the cap every month, your volume has outgrown a demo allowance, and the conversation to have is about client builds, where voice is billed per minute anyway.
Does chatting with the agent use voice minutes?
No. Chat draws on the 500 message count, voice on the 100 minute count. Tuning a script in chat costs you nothing in minutes, which makes it the cheapest way to iterate.
What happens to my limits if I cancel?
Published demos unpublish so the links stop serving, drafts stay exactly as they are, and you keep your login with zero voice, chat, and audits until you resubscribe. Nothing is deleted. The site FAQ says the same in fewer words.
Do the arithmetic on your own pipeline first
Count the demo links you would send in a month. Multiply by the share of prospects who open and talk. Multiply that by three minutes. Under 100 and minutes are not your problem, so spend your thinking on the one published demo cadence instead. Over 100 and you should say your volume out loud on the pricing call, and we will tell you whether this is the right plan for you.
Talk to the demo receptionist on the home page first, since that is the thing your prospects will spend your minutes on, then start an agency workspace when the arithmetic works.
