FieldCampSet your voice agent's language | FieldCamp
Set the language your FieldCamp voice agent listens for, use auto-detect when callers speak more than one language, and match voices, prompts and audio clips.
A voice agent's language affects three things: what it listens for, what it says, and how it sounds. In FieldCamp Voice, each is set in a different place. This page explains where, and how to set up an agent for callers who don't all speak English.
Where language is set
| What | Where | What it controls |
|---|---|---|
| Listening language | Models page, Language setting (when your model offers one) | The language the agent expects to hear |
| Reply language | Your agent's prompts | The language the agent answers in |
| Voice | Models page, Voice section | How the agent sounds in that language |
| Audio clips | Audio Clips, Language setting on upload | How each clip is transcribed |
The listening language is a workspace setting, like the rest of the models and voices. An agent with its own model override uses the language in its override.
Set the listening language
Where the Language setting appears depends on how your calls run.
- Model and voice: a Language section appears below Transcription when the transcription model you picked offers a language choice. The options come from that model. If your transcription model doesn't show a Language section, there's no listening language to set on the Models page.
- Speech-to-speech: the Language setting sits below the voice. It includes Auto-detect, and you can pick one language instead.
To set it, open Models under Configuration, choose an option in Language, and click Save changes. The language options are the ones FieldCamp offers for your model, so what you see can differ from another workspace.
Auto-detect
With auto-detect, the agent works out which language the caller is speaking on each call.
- In Model and voice mode, the option is Multilingual (Auto-detect). Its help text reads "The agent detects the language on each call."
- In Speech-to-speech mode, the option is Auto-detect. Its help text reads "The model detects the language on each call. Choose one if your callers only speak it."
If all your callers speak one language, pick that language instead. Setting it helps the model hear callers accurately.
Multilingual callers
Many transcription and speech-to-speech models include Spanish in their language options, but the options you see depend on the models FieldCamp offers in your workspace. For a plumbing or HVAC line that takes calls in English and Spanish, set the listening language to auto-detect so the agent can handle either.
Hearing a language isn't the same as answering in it. Tell the agent what to do in your prompts, for example:
If the caller speaks Spanish, reply in Spanish for the rest of the call. Read back the service address in the same language the caller used.
Put this in the agent's persona so it applies to every step. See Write voice agent prompts for field service for where persona and rules go.
Voices and prompts in other languages
A voice that sounds natural in English may not sound natural in another language. On the Models page, use the Language filter in the voice list to narrow the voices, and preview each one in the language your callers speak. For ElevenLabs voices, the Voice model list includes multilingual models such as eleven_multilingual_v2.
You can write prompts in English and tell the agent which language to reply in, as in the example above. Keep fixed wording, such as your company name and service area names, spelled the way you want them said.
Pre-recorded audio clips have their own Language setting, which defaults to Multilingual (Auto-detect). Set it before you add files, and record a separate clip for each language your agent answers in.
Test each language
Test every language you plan to support before you publish. Talk to the agent in the language with Test Audio, then place a test call to your phone to hear it on a real line. Check that it:
- Understands names, street names and trade terms
- Replies in the language you told it to use
- Reads back addresses and phone numbers correctly
- Answers questions from your business knowledge correctly in that language
- Hands off or ends the call in the same language
Frequently asked questions
Can an AI receptionist speak Spanish?
Many transcription and speech-to-speech models include Spanish, depending on the models offered in your workspace. To have the agent answer in Spanish, set the listening language, tell the agent in its prompts when to reply in Spanish, and pick a voice that sounds natural in Spanish. Test it before you publish.
How does automatic language detection work?
With Multilingual (Auto-detect) or Auto-detect selected, the agent works out which language the caller is speaking on each call. If your callers only speak one language, choose that language instead so the model hears them more accurately.
What happens when a caller switches languages?
Auto-detect is described as detecting the language on each call, so don't rely on it to follow a caller who switches partway through. Tell the agent in its prompts which language to use, and test a call where the caller switches to see how your setup handles it.
Do prompts need to be in the caller's language?
No. You can write prompts in English and tell the agent which language to reply in. Write fixed wording, such as your company name, the way you want it said.
Does the voice change with the language?
No. The agent uses the voice you selected on the Models page for every call. Pick a voice that sounds natural in each language you support, and preview it in that language.
Related articles
Choose a model and voice for your voice agent | FieldCamp
Choose how your FieldCamp voice agents run calls, pick the language model, voice and transcription model, set a fallback voice, and override one agent's models.
Tune turn-taking and interruptions | FieldCamp
Control how long your voice agent waits before replying, how easily callers can interrupt, backchannel acknowledgments, idle reminders, and silence hang-ups.