Talking Avatar
Pricing research
AI Voice Agent Pricing in 2026: What Retell, Vapi, Bland and ElevenLabs Cost per Minute
By Mukul Munjal, Founder, InterviewFlowAI8 min read
AI voice agents cost $0.06–$0.11 a minute at the cheapest published self-serve rate, from each vendor's own pricing page on 24 October 2026: Cartesia $0.06, Retell AI from $0.07, ElevenLabs and Vapi about $0.08, and Bland AI $0.11–0.14. Those rates rarely include everything: Retell, Vapi and ElevenLabs bill the LLM separately, most charge for telephony, and Retell's own all-in range goes up to $0.31. Agents with a realistic avatar cost $0.11–$0.50 a minute from Anam, HeyGen, Tavus and D-ID, and $0.06 on Talking Avatar Realistic Avatars.
- lowest published rate per minute across voice-only agent platforms
- $0.06–$0.11
- top of Retell AI's own all-in range per minute, with a premium LLM and voice
- $0.31
- per minute for managed agents with an avatar from other vendors
- $0.11–$0.50
- per minute for Realistic Avatars voice agents, avatar, voice and AI included
- $0.06
Key takeaways
- The headline rate is a floor. Retell ($0.055) and Vapi ($0.05) quote a platform fee, and the speech, LLM, voice and phone line come on top.
- Bland AI and Cartesia bundle the most into one rate, but both still bill telephony separately, and Cartesia's free LLM is a limited-time offer.
- Concurrency costs extra once you grow: Retell charges $8 and Vapi $10 per additional line a month, and ElevenLabs doubles the rate for calls over your limit.
- Synthflow no longer publishes self-serve prices, and OpenAI prices its Realtime API in tokens, not minutes.
- A face costs more on most platforms: managed avatar agents run $0.11–$0.50 a minute. Talking Avatar Realistic Avatars voice agents are $0.06 with the avatar, for website calls rather than phone calls.
Price per minute, cheapest first
Lowest published self-serve price per minute (USD)
Voice agents checked 24 October 2026; avatar agents October 2026, from each vendor's own pricing page. What each rate includes differs: see the table below. Hover a row for its type.
- Talking Avatar Realistic Avatars voice agents$0.06
- Cartesia agents$0.06
- Retell AI$0.07
- ElevenLabs Agents$0.08
- Vapi$0.08
- Bland AI$0.11
- Anam$0.11
- HeyGen LiveAvatar (FULL)$0.158
- Akool (Streaming Avatar)$0.17
- Akapulu Labs$0.175
- Runway Characters$0.20
- Tavus (CVI)$0.244
- D-ID Agents$0.35
- Microsoft Azure TTS Avatar (real time)$0.50
Log scale: each step to the right is a multiple of the price, not a fixed amount.
What each voice agent platform charges
| Platform | Published price | What it includes | Free tier, billing and limits |
|---|---|---|---|
| Talking Avatar Realistic Avatars voice agents | $0.06/min, billed per second | Avatar, voice and AI; web only | No subscription; $1 of free credits; calls up to 60 minutes; share link, widget or iframe, no phone numbers |
| Cartesia agents | $0.06/min, plus $0.014/min telephony | Call minute; LLM free for UI-built agents for a limited time; telephony extra | Free plan includes $1 of agent credit a month; paid plans from $5/month as prepaid credit; 8 to 20 concurrent |
| Retell AI | $0.07–0.31/min all-in (Retell's own range) | Voice infrastructure $0.055/min, plus LLM, voice and telephony metered separately | $10 of free credits, no subscription; tracked to the nearest second; 20 concurrent calls free, then $8 a line a month |
| ElevenLabs Agents | $0.08/min, plus LLM | Call minute; LLM billed separately | Free plan has 15 minutes; paid plans from $6/month; calls over the concurrency limit cost double |
| Vapi | $0.05/min hosting; about $0.08–0.13/min all-in in Vapi's calculator | Hosting $0.05/min, plus speech, LLM, voice and telephony passed through | $5 of free credits; 4 concurrent without a plan, 10 on Core ($29/month), then $10 a line |
| Bland AI | $0.14/min with no platform fee; $0.12 on Build ($299/month); $0.11 on Scale ($499/month, in its billing docs) | Speech, LLM and voice bundled; telephony extra | Prorated to the second; $0.015 minimum per outbound attempt or failed call on Bland's telephony; 10 concurrent on Start, 50 on Build |
| Synthflow | Enterprise only; contracts from $30,000 a year | Not published | Self-serve pay-as-you-go is no longer listed on its pricing page |
| OpenAI Realtime API | gpt-realtime: $32 per 1M audio input tokens, $64 per 1M audio output tokens | Tokens only: you add your own hosting and telephony | OpenAI publishes no per-minute figure for its realtime models; the mini model is $10 and $20 |
The costs on top of the per-minute rate
- The LLM. Retell lists models from $0.0016 to $0.64 a minute, and ElevenLabs bills LLM usage on top of its $0.08. The model you pick can cost more than the platform.
- The voice. On Retell, its own voices add $0.015 a minute and ElevenLabs voices $0.04–0.10.
- The phone line. Telephony runs about $0.008–0.015 a minute (Twilio rates on Vapi, $0.015 on Retell, $0.014 on Cartesia), plus a monthly fee per number.
- Concurrency. Retell includes 20 calls at once, Vapi 4 without a plan, Bland 10 on Start; more lines are a monthly fee or a bigger plan.
- Minimums. Bland charges at least $0.015 per outbound attempt or failed call on its own telephony. Among avatar agents, Tavus bills a 30-second minimum per call.
What 1,000 minutes a month costs
At 1,000 minutes, the published numbers give roughly: Cartesia $60 plus $14 of telephony; Retell $70–310 depending on the LLM and voice; Vapi $82–129 by its own calculator; ElevenLabs $80 plus the LLM; and Bland $140 on Start, plus telephony. With an avatar, Talking Avatar Realistic Avatars voice agents are $60 with nothing billed on top, and the cheapest managed avatar agent we found, Anam, starts at about $110 in overage alone, before its plan fee.
Voice only, or a voice agent with a face?
The platforms above are built for phone calls: inbound support lines, appointment booking and outbound dialling. Several also offer web calls and widgets. If your callers dial a number, they're the right comparison, and Talking Avatar Realistic Avatars voice agents aren't: they don't answer phone numbers.
If the conversation happens on your website (support, sales, interviews, tutoring), a realistic face changes the experience. Talking Avatar Realistic Avatars voice agents put a photorealistic avatar on a voice agent you build without code, for $0.06 a minute, billed per second, with the avatar, voice and AI included. You share them as a link, a corner widget or an iframe.
How to choose on price
- Phone calls at volume: compare all-in cost per minute (platform, LLM, voice, telephony) at your expected concurrency, not the headline rate.
- You want one predictable rate: Bland and Cartesia bundle the most; check what telephony adds.
- You want to pick every component: Vapi and Retell let you choose the speech, LLM and voice providers and pay their rates directly.
- Your callers are on your website: an agent with an avatar is worth pricing. Talking Avatar Realistic Avatars voice agents are $0.06 a minute all-in; managed avatar agents elsewhere start at $0.11.
- You already run a voice agent: add a face through an avatar API instead. Talking Avatar Realistic Avatars is $0.02 a minute; see our avatar API price comparison.
Methodology
Voice agent prices come from each vendor's own pricing page or docs, read on 24 October 2026; avatar agent prices come from our avatar API pricing table, checked on October 2026. The "lowest published self-serve price" is the cheapest per-minute rate available without an Enterprise contract. Where a vendor quotes a platform fee plus pass-through costs, we show its own published all-in figure (Retell's range, Vapi's calculator). Vapi's per-minute figure is its calculator total divided by 1,000 minutes. Prices change often; follow the sources to check today's. We build Talking Avatar, and we've listed where others fit better.
Frequently asked questions
How much does an AI voice agent cost per minute?
Self-serve voice agent platforms publish rates from $0.06 to $0.11 a minute for their cheapest option, and all-in costs reach $0.31 a minute on Retell AI with a premium LLM and voice. Telephony, LLM usage and extra concurrent lines are often billed on top. Agents with a realistic avatar cost $0.11–$0.50 a minute from other vendors, and $0.06 on Talking Avatar Realistic Avatars.
What does Retell AI cost?
Retell AI charges $0.055 a minute for its voice infrastructure, plus the LLM, voice and telephony you choose, for $0.07–0.31 a minute all-in by its own range. New accounts get $10 of free credits with no subscription, calls are tracked to the nearest second, and 20 concurrent calls are free before $8 a line a month.
What does Vapi cost?
Vapi charges $0.05 a minute for hosting and passes through the speech, LLM, voice and telephony providers you pick. Its own calculator puts 1,000 minutes at about $82–129, roughly $0.08–0.13 a minute. You get $5 of free credits; without a plan you can run 4 calls at once, or 10 on the $29-a-month Core plan.
What does Bland AI cost?
Bland AI's Start plan has no platform fee and charges $0.14 a minute with speech, LLM and voice included; telephony is billed separately. Paid plans lower it to $0.12 ($299 a month) and $0.11 ($499 a month). Calls are prorated to the second, with a $0.015 minimum per outbound attempt or failed call on Bland's telephony.
Does Synthflow publish pricing?
Not any more. On 24 October 2026 its pricing page listed only Enterprise plans, with contracts from $30,000 a year, and its pay-as-you-go page redirected there.
How much does the OpenAI Realtime API cost per minute?
OpenAI prices its realtime models by tokens, not minutes: gpt-realtime is $32 per million audio input tokens and $64 per million audio output tokens, and the mini model is $10 and $20. It doesn't publish a per-minute figure, and you add your own hosting and telephony.
Why do voice agents with an avatar cost more?
They render a video of a person in real time, on GPUs, on top of the speech, LLM and voice. Managed avatar agents from other vendors cost $0.11–$0.50 a minute. Talking Avatar Realistic Avatars voice agents cost $0.06 a minute with the avatar included, about the same as voice-only platforms' cheapest rates.