Voice AI Stack Atlas

The voice AI stack, mapped.

Pick a layer, see who builds there, click through to a template that runs it in production.

Telephony

Daily

WebRTC and SIP infrastructure for real-time voice and video. Hosted.

$$Pay-as-you-go per participant minute1 template

Plivo

Voice and messaging CPaaS with global PSTN coverage. Hosted.

$$Pay-as-you-go per minute plus number rental5 templates

Telnyx

Carrier-grade voice, SIP, and messaging on a private IP backbone. Hosted.

$Pay-as-you-go per minute, by destinationNo templates yet

Twilio

Programmable voice, SMS, and SIP APIs. Hosted.

$$$Pay-as-you-go per minute, with Media Streams add-on33 templates

Vonage

Voice, video, and messaging APIs from the Vonage Communications Platform. Hosted.

$$$Pay-as-you-go per minute, by destination3 templates

Speech-to-text

AssemblyAI

Speech-to-text API with streaming and async transcription. Hosted.

$$Pay-as-you-go per second of audio10 templates

Deepgram

Speech-to-text API. Hosted and self-hosted.

$Pay-as-you-go per minute, plus on-prem option64 templates

Google Speech

Speech-to-Text API on Google Cloud with streaming and batch modes. Hosted.

$$$Pay-as-you-go in 15-second increments4 templates

OpenAI Whisper

Speech-to-text models from OpenAI, available via API or as open weights. Hosted and self-hosted.

$$Pay-as-you-go per minute, or run open weights yourself5 templates

Speechmatics

Speech-to-text API and on-prem container with streaming and batch modes. Hosted and self-hosted.

$$Pay-as-you-go per hour of audio, on-prem licensedNo templates yet

LLM

Anthropic Claude

Claude family of language models from Anthropic. Hosted.

$$$Pay-as-you-go per million tokens, by tier13 templates

Google Gemini

Gemini family of language models from Google. Hosted.

$$Pay-as-you-go per million tokens, audio billed separately14 templates

Groq

Inference API serving open-weight models on LPU hardware. Hosted.

$Pay-as-you-go per million tokens, by model3 templates

Mistral

Open-weight and hosted language models from Mistral AI. Hosted and self-hosted.

$$Pay-as-you-go per million tokens, weights free to self-hostNo templates yet

OpenAI GPT

GPT family of language models from OpenAI. Hosted.

$$$Pay-as-you-go per million tokens, Realtime billed per minute60 templates

Text-to-speech

Azure TTS

Neural text-to-speech voices from Azure AI Speech. Hosted.

$$Pay-as-you-go per million characters2 templates

Cartesia

Streaming text-to-speech built on state space models. Hosted.

$$Subscription tiers metered by characters34 templates

ElevenLabs

Text-to-speech platform. Hosted.

$$$Monthly tiers metered by characters24 templates

OpenAI TTS

Text-to-speech voices from OpenAI, available via API. Hosted.

$$Pay-as-you-go per million characters7 templates

PlayHT

Text-to-speech and voice cloning APIs. Hosted.

$$Monthly tiers, with API plans on top1 template

Resemble AI

Text-to-speech and voice cloning with hosted and on-prem options. Hosted and self-hosted.

$$$Monthly tiers, with on-prem quoted separatelyNo templates yet

Orchestration

Bland

Voice agent platform with bundled telephony and models. Hosted.

$$$Flat per-minute rate, all-in3 templates

LiveKit

Open-source realtime infrastructure with an Agents framework for voice. Hosted and self-hosted.

$$Cloud metered by participant minutes, server is free14 templates

Pipecat

Open-source Python framework for realtime voice and multimodal agents. Self-hosted.

$Framework is free, you pay underlying vendors23 templates

Retell

Voice agent platform with model routing and telephony integrations. Hosted.

$$$Per-minute platform fee, premium voices extra1 template

Vapi

Orchestration platform for voice agents. Hosted.

$$$Per-minute platform fee, plus vendor pass-through5 templates

Vocode

Open-source Python library for building voice agents. Hosted and self-hosted.

$Library is free, hosted tier on request1 template