Japanese · 日本語

An AI voice agent that speaks Japanese.

A Japanese agent writes each reply in kanji and kana before it is spoken in a Cartesia Sonic voice, so the voice reads Japanese as it is written rather than from romaji. GPT Transcribe hears the caller.

What a caller hears.

A greeting as the agent writes it, in kanji and kana, before the voice speaks it.

こんにちは!本日はどのようなご用件でしょうか?

Hello! How can I help you today?

Voices that speak it.

Choose the engine, then the voice; listen to a sample before you pick.

  • Cartesia Sonic 3.6. Speaks every language on the list, with speed and tone controls.

Engines that hear it.

What turns the caller’s speech into text for the agent to read.

  • GPT Transcribe. From OpenAI, in every language.

Good to know about Japanese.

  • The register is set in the instructions: ask for keigo for a customer desk, or plain speech for a companion app, and the agent keeps to it.
  • Start one call in Japanese with language: "ja" from an English agent with a Cartesia voice, without changing the agent.
  • English words in a Japanese sentence are understood: every agent listens for English too.

Start a call in Japanese.

Save the agent in Japanese in the studio, or keep it as it is and hold one call in Japanese by sending its code when the call starts.

One call in Japanese

curl -X POST "https://goblins-call-api.onrender.com/v1/calls" \
  -H "Authorization: Bearer $CRACKLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "agentId": "ua_3f9a1c2b", "language": "ja" }'

The agent’s voice must be Cartesia Sonic; any other voice is answered 400 UNSUPPORTED_LANGUAGE. GPT Transcribe hears the caller; an agent saved with Sarvam Saaras is transcribed by GPT Transcribe for this call, with no change to the agent. This applies only to a call held in a language other than the agent’s own.

Questions.

Can a Japanese caller use English words?

Yes. Every agent listens for English as well as its own language, so a sentence that mixes the two is understood.

Can one agent take calls in Japanese and in other languages?

Yes, in two ways. Save the agent in Japanese and every call starts in it. Or keep the agent as it is and hold one call in Japanese by sending language: "ja" when you start it. The agent’s voice must be Cartesia Sonic; any other voice is answered 400 UNSUPPORTED_LANGUAGE. GPT Transcribe hears the caller; an agent saved with Sarvam Saaras is transcribed by GPT Transcribe for this call, with no change to the agent. This applies only to a call held in a language other than the agent’s own.

What does a call in Japanese cost?

The same as in any other language: $0.07 a minute, whichever voice and model, billed by the second.

Other languages.

Give your agent a voice in Japanese.

Build one in the studio and talk to it in your browser. Sign-up credit is there for your first test calls.