Simple Plans for
Scaleable Voice AI
We separate infrastructure costs from engineering services. Pay for the setup to get it right, then pay for usage to scale it up.
Basic
For pilots and initial deployment.
-
0.80 cents per minute
-
2 Months Optimization Free
-
1 Custom Agent Config
-
Standard Email Support
Pro
For scaling production workloads.
-
0.75 cents per minute
-
4 Months Optimization Free
-
1 Custom Agent Config
-
Priority Support Channels
Enterprise
Custom solutions for high volume.
-
Lowest per-minute rates
-
Dedicated Support Engineer
-
Unlimited Concurrency
-
Custom LLM Fine-tuning
What goes into your plan?
We aren't just selling API keys. We are selling a fully managed voice infrastructure layer.
The Setup Fee
Voice AI requires specialized tuning. The fee covers 20+ hours of dedicated engineering to:
- Fine-tune turn-taking latency sensitivity.
- Configure SIP trunking for your region.
- Stress-test LLM prompts for voice edge cases.
Free Improvements
Included in your plan is Human-in-the-loop Optimization. We proactively monitor your calls.
- Basic: 2 Months of weekly log analysis.
- Pro: 4 Months of deep-dive analysis.
- We tweak VAD settings to stop interruptions.
Usage Rates
Your per-minute rate covers the entire technology stack required to hold a conversation:
- Telephony: Carrier costs & number rental.
- Transcriptions: Deepgram/Nova-2 STT.
- Orchestration: The WebSocket latency engine.
Frequently Asked Questions
Billing & Costs
How is usage calculated?
We bill per minute of connection time. Billing starts when the WebSocket connection is established (or the call is answered) and ends when the connection closes. We bill in 1-second increments with a 15-second minimum per call.
Are there extra costs for telephony?
No. Your per-minute rate includes the telephony leg (SIP/PSTN), the phone number rental, and the transcription costs. The only extra cost is the LLM tokens, which you pay directly to OpenAI/Anthropic via your own API key.
Can I cancel anytime?
Yes. The monthly fee is charged at the start of the billing cycle. You can cancel at any time, and your service will continue until the end of that billing period.
Technical Capabilities
Which LLMs do you support?
We are LLM agnostic. You can plug in OpenAI (GPT-4o, GPT-3.5-Turbo), Anthropic (Claude 3.5 Sonnet, Haiku), or any open-source model hosted on Groq or Together AI. We just need the API key.
Can I use my own Twilio account?
On the **Enterprise** plan, we support BYOC (Bring Your Own Carrier). On Basic and Pro plans, we manage the telephony carriers to ensure quality of service and low jitter.
What is the latency?
Our average "Time to First Byte" (TTFB) is ~600ms. This includes transcription, LLM inference (using fast models like Groq), and TTS generation. Our turn-taking engine makes this feel instant to the user.
Security & Compliance
Do you store call recordings?
By default, we store logs and recordings for 30 days to help you debug. You can configure your account to delete recordings immediately after the call is finished for strict privacy compliance.
Is Invaria HIPAA compliant?
Yes, we can sign a BAA (Business Associate Agreement) for customers on the **Pro** and **Enterprise** plans. Our infrastructure is SOC2 Type II compliant.