Developers building phone agents that keep mangling phone numbers and interrupting callers; the free plan gives nearly two hours a month to test.
Kenning's take
What it does
Ink-2 turns live speech into text for voice agents, such as phone support bots. It is built to handle phone numbers, dates and IDs correctly and signals when a caller starts and stops talking, so an agent does not cut in mid-sentence. Developers use it through the Cartesia API, and it now supports English, Spanish, French, Hindi and Japanese.
- Streaming transcription for real-time voice agents
- Native turn detection with turn.start and turn.end signals
- Semantic endpointing based on meaning, not silence
- Handles phone numbers, dates, emails and IDs
- English, Spanish, French, Hindi and Japanese
Who it suits. Developers building phone and voice agents that need fast, accurate live transcripts.
Key facts
- Category
- AI models
- Made by
- Cartesia
- Based in
- United States
- Released
- 9 July 2026
- How to get it
- through the Cartesia API and playground, on every Cartesia plan including the free one
- Works on
- API, Web playground
- Free plan
- Free plan, 20K credits a month (about 1 hour 51 minutes of Ink-2 transcription)
- API price
- credit-based; Pro's 100K credits a month cover about 9h 16m of Ink-2
Pricing
| Plan | Price | Billed |
|---|---|---|
| Pro | $5 | month |
| Startup | $49 | month |
| Scale | $299 | month |
| Enterprise | Custom pricing | custom |
Reviews and write-ups
- RuntimeWire 2026-06-15
FAQ
When was Ink-2 released?
Cartesia published its Ink-2 announcement on 9 July 2026.
Is Ink-2 free?
Cartesia's free plan includes 20K credits a month, which covers about 1 hour 51 minutes of Ink-2. Paid plans start at $5 a month.
What languages does Ink-2 support?
English, Spanish, French, Hindi and Japanese, per Cartesia's docs.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.