aethexDocs
Dashboard

Languages

Aethex agents can speak and understand more than one language. You pick the language when you create the agent, and everything (listening, replying, and the voice) uses it.

Supported languages

Each supported language runs on speech models trained on recordings of native speakers, so agents sound natural rather than translated. Today you can choose:- English (english)
- French (french)More languages are on the way.

Setting the language

Set the language with the language field when you create or update an agent. Use the lowercase name, for example english or french. The example below creates a French support agent.

Note
Voices are tied to one language. Before you set voice_id on a French agent, list the French voices with GET /voices?language=french and pick one of those. An English voice on a French agent will sound wrong.
Create a French-language agent
bash
curl -X POST https://api.aethexai.com/api/v1/agents \
  -H "X-API-Key: $AETHEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Support FR",
    "system_prompt": "Tu es un agent du service client.",
    "voice_id": "5c34046a-ac9b-57d5-8c70-5a61e694be3f",
    "language": "french"
  }'

What changes with the language

Setting language is all you need. From that one field, the agent listens for that language, thinks and replies in it, and speaks it with a native-sounding voice. You do not have to configure anything else for the language to take effect.

Dialect style

dialect_style controls how far the agent leans into a local way of speaking, such as Nigerian Pidgin on an English agent. It takes one of three values:- `formal` (the default): standard, neutral delivery.
- `mixed`: mostly standard, but happy to slip into the local variety when the caller does.
- `local`: leans fully into the local variety.The local accent only comes through if the agent's voice supports it. Pick a voice with supports_dialect_style: true (see the Voices page). Leave dialect_style out to use formal.

standardfull localformalmixedlocaldefaultThe local accent needs a voice with supports_dialect_style: true.

Next

Pick a voice for your language: Voices.