Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cartesia.ai/llms.txt

Use this file to discover all available pages before exploring further.

Cartesia’s models can be self-hosted into customer provisioned cloud environments, such as GCP, AWS, or on-premise data centers.

Why Self-Host

Cartesia’s public API is globally available for the lowest latency, complete with GDPR, SOC 2 Type II, PCI Level 1, and HIPAA compliance with enterprise contract options for Service Level Agreements (SLA) and Business Associate Agreement (BAA), and more. However certain use cases may still warrant Self-Hosted Voice AI and Cartesia supports both private cloud and on-premise hosting options. In those circumstances we recommend a self-hosted offering that is feature complete and as performant as the cloud offering.

Colocation

With self-hosted deployments, you can choose to colocate your Voice AI models with other offerings and establish your own SLAs around uptime and throughput. Colocated TTS would save a lot on network latencies depending on where your datacenters are located.

Isolation (Single Tenant)

Even though we provide a tenant level isolation in our cloud offering, nothing will beat the isolation you can achieve by self-hosting.

Security

Self-hosted deployments allow you to maintain tight security posture without running Voice AI traffic over the internet to our public APIs. The self-hosted deployments will only contact the Cartesia server to authenticate model access and report usage information. Usage information is limited to metadata such as character count and voice id, and does not contain any transcript information. We also support air-gapped deployments where there’s no contact to our cloud, instead your deployment works with an offline license.

Sovereignty

You can choose to host your Voice AI offering in any geographic region with GPU availability to meet jurisdictional requirements.

Supported Products

ProductSupport
Sonic 2Kubernetes
Sonic 3Kubernetes, AWS SageMaker
Ink WhisperKubernetes
Voice AgentsNot supported
Voice CloningNot supported