Skip to main content
Smart Translation turns a chat message into another language while keeping the meaning, tone, and context of the conversation intact.

​
How It Works

Smart Translation is built for two-sided chat between a User (your subscriber) and a Persona (your animator). Instead of translating a single line in isolation, it reads the surrounding conversation and takes each side into account, so the result reads naturally in both directions.

​
Key Benefits

  • Context Aware: The whole conversation is used as context, so short or ambiguous replies (like “sure”, “him”, or “over there”) are translated correctly.
  • Gender Aware: You tell us each side’s gender, so grammar, agreement, and forms of address come out right in every language.
  • Two-Sided: A dedicated endpoint for each side of the chat keeps User and Persona messages accurate in both directions.
  • Pay Per Use: Each translation is billed individually and appears as its own line item in your statistics.

​
Common Use Cases

  • Letting a User and a Persona chat naturally across two different languages.
  • Localizing an ongoing conversation for a human animator working in another language.
  • Translating incoming messages before they reach your moderation or support teams.

​
Endpoints

There are two endpoints, one for each side of the conversation:

​
Making a Request

​
What You Provide

  • Source and target language: the language to translate from and the one to translate into. See Supported Languages.
  • The conversation so far: the recent messages, oldest first, so we understand the context.
  • User and Persona gender: used to get pronouns, agreement, and tone right.

​
How to Translate a Message

  1. Pick the endpoint Use User Message Translation for a User’s message, or Persona Message Translation for a Persona’s message.
  2. Set the languages Choose the source language and the target language for the translation.
  3. Send the conversation Include the recent messages, oldest first. The last message must belong to the side you chose in step 1.
  4. Add the genders Provide the User’s and Persona’s gender so the translation reads naturally.
  5. Read the response You get back the translated message and the cost of that translation.
Important: Only the most recent message is translated — the earlier messages are used purely as context. The last message must be from the side matching the endpoint you called (User for /user, Persona for /persona).