Put a dollar figure on every Cartesia credit, by API key, model, and owner

Your Cartesia voice AI spend now shows up in Cloud Intelligence™, priced from your plan and traced to the keys, models, and people behind it. It sits next to the rest of your cloud and AI bill.

Cartesia meters text to speech, speech to text, infill, and voice cloning in credits, and credits carry no dollar price. The usage page tells you how many credits each capability consumed. It doesn't tell you which service, which model version, or whose API key drove the increase. Hosted voice agents bill separately in dollars per minute, so the two halves of your Cartesia bill don't even share a unit. If the monthly allotment is gone by week three, working out why means exports and guesswork.

Cloud Intelligence™ now turns each Cartesia credit into a cost you can put a name on. You see the key that made the request, the model version that served it, and who created the key, priced from your plan.

What you get

Dollars without a rate card. Choose your subscription plan when you connect, and credits are priced at the plan's implied rate: the monthly price divided by the included credits. On the Startup plan, $49 for 1.25M credits comes out to about $0.039 per 1,000 credits. You don't type in a price per 1,000 credits or keep a rate schedule current. No months of history sit at zero cost because someone forgot a date range. Agent minutes arrive as billed in US dollars, so credits and agents finally add up to one number.

Attribution past the capability. Many cost views stop at "Text to Speech" as a single line. Here the full capability total is split by API key, model version, and key owner email. Playground usage, unnamed keys, and deleted keys each get their own attribution label, so the long tail stays visible instead of disappearing into "other". Allocate by key owner and voice AI cost rolls up to the teams that actually ship it.

Totals that match Cartesia. Credits that Cartesia reports but doesn't tie to any key are labeled unattributed rather than dropped. Your report total matches Cartesia's usage page. We checked the breakdowns against live usage data before shipping.

Model versions you can compare. Each model version, such as sonic-3-2026-01-12 or ink-whisper-2025-06-04, is its own SKU. When a team moves to a new Sonic release, you can see the spend shift in the same report.

The tooling you already use. Cartesia data lands in Cloud Analytics with your other providers and appears in the GenAI Intelligence dashboard next to your other AI spend. Budgets, anomaly detection, and allocations cover voice AI the same way they cover everything else. Up to 12 months of history is imported when you connect.

Get started

Setup is deliberately boring.

1. Create an admin API key. Sign in to the Cartesia Playground as an organization admin and create an admin key. Admin keys can read usage but can't generate speech or transcribe audio, so the key you share has no access to your voice workloads. Create an admin API key
2. Connect your organization. In Cloud Intelligence™, go to Data ingestion and integrations > Integrations, search for Cartesia, and select Connect +. Enter the organization name, choose the plan that matches your subscription, paste the key, and test the connection. Enterprise plans have no public price, so those credits show as usage until your negotiated rate is set [CONFIRM: how Enterprise customers set their rate]. Cartesia connection guide
3. Let the history load. The initial import takes a few hours, and you'll get an email when the data is ready in Cloud Analytics. How credits become cost
4. Break it down. Open a Cloud Analytics report on the Cartesia provider and group by the API key owner or Model label to see where credits go. Cartesia reporting dimensions