Voice platform
Phonic
Speech-to-speech voice agent platform with one per-minute price that includes the model.
- Pricing at a glance
- $0.12-$0.14/min conversation; free Starter includes $30/mo credits
Official plans page; the free Starter plan includes $30 of usage credit each month.
Use one speech-to-speech model at one per-minute price
What you get
Build agents on an end-to-end speech-to-speech model with tool calling, transfers and SIP or LiveKit connections, priced per conversation minute.
What to account for
You cannot swap individual speech or LLM components, latency claims are vendor-reported, and carrier costs sit outside the rate.
What the platform provides
- Speech-to-speech agents
- An end-to-end speech model with LLM supervisors for tool calls, built in the dashboard, SDKs, webhooks or LiveKit.
- Phone connections
- SIP trunks such as Twilio and Telnyx, or Amazon Connect, with transfer tools.
Connections & compatibility
- Connects with
- Twilio SIP, Telnyx SIP, Amazon Connect, LiveKit, Webhooks, MCP
Cost at your volume
Starter is free with $30 of monthly credit at $0.14 per conversation minute and 10 concurrent calls; Pro is $240 a month including $240 of credit at $0.12/min and 50 concurrent calls. Phonic says there are no LLM pass-through or per-token charges. SIP carrier and number costs are separate, and the docs cite a $0.15/min starting rate, so confirm the current price.
Before committing
Interrupt the agent mid-sentence on a noisy line, fail a tool call, then transfer. Check what the transcript and analysis show afterwards.
- What response latency do you measure over our SIP trunk and region?
- Which parts of a call are handled by the speech model versus the supervising LLM, and what do we see in logs?
- How does transfer behave when the destination does not answer?
- Can you share the SOC 2 report and BAA, and where are recordings and transcripts stored?
Controls to confirm
Phonic states SOC 2 Type II, HIPAA with a BAA, SSO, a 99.99% uptime SLA and single-tenant, VPC or self-hosted deployment on Enterprise. Request the SOC 2 report and BAA, and confirm recording retention, which LLM providers see transcripts, and tool permissions. Morak has not audited these claims.
What to consider next
A different approach
Compare OpenAI's Realtime API if you want a speech-to-speech model you orchestrate yourself.
A different approach
Consider Vapi if you want to pick each speech and LLM provider separately.