Cartesia

The October 2026 EliseAI case describes Sonic voice-model use for more than five million monthly calls, regional voice requirements and gated evaluation before broader production rollout.

Google Cloud's Cartesia case describes production speech generation handling millions of requests per day with GKE-based infrastructure and scaling.

Cartesia provides low-latency voice intelligence APIs used in real-time AI applications, with growing production adoption.

Cartesia states its founding researchers met at Stanford AI Lab and invented state space model architectures, subsequently extending their research across text, audio, video, images and time series. This is a named technical lineage separate from the generic description of current real-time voice models.

Cartesia reports rebuilding its data, model architecture, training and evaluation for Sonic-3.6 and reports listener preference of up to 93% in blind head-to-head tests across fifteen locales compared with Sonic-3.5.

Cartesia develops state-space-model-based real-time voice and multimodal intelligence systems.

Cartesia develops state-space-model-based real-time voice and multimodal intelligence systems.

Cartesia describes an in-person organization with offices in San Francisco, London and Bangalore, and publishes medical coverage, parental leave, retirement-plan and relocation support, adding specific organization and employment-policy evidence beyond its ServiceNow integration.

Cartesia's official customer page identifies integration of its voice AI technology with ServiceNow AI Voice Agents for secure, real-time enterprise use. The named enterprise integration is independent of its existing generic developer-platform evidence.

Cartesia operates a developer-facing real-time inference platform and model engineering organization.

Cartesia's historical Python SDK release documents asynchronous requests, streamed responses, WebSocket integration, automatic exponential-backoff retries, configurable timeouts and custom HTTP transports.

Inference, speech, state-space modeling and systems skills transfer broadly across real-time AI products.

Inference, speech, state-space modeling and systems skills transfer broadly across real-time AI products.