Speako's AI agents speak more than 30 languages, but how the agent handles each language depends on which ones you've enabled and whether your callers may switch between them. This article explains what your agent can and cannot do across languages, so you can configure the right setup for your business.
The Two Language Modes
Every Speako agent runs in one of two modes during a call, and which one applies depends on how you've set up languages in Sidebar → AI Frontdesk → Voice & Languages.
Single-Language Mode
If you've only set a Primary Language, the agent stays in that language for the entire call. There is no language switching, and any callers who speak a different language will hear the agent continue in the primary language.
This mode is ideal if all your callers speak the same language, and it has the broadest language coverage — the primary language can be any of the 30+ languages listed in the language picker.
Multi-Language Mode
If you add additional languages and turn on the Enable additional languages master switch, the agent can detect the caller's language and switch to it mid-call. This mode unlocks real-time language detection, but the available languages are more limited than in single-language mode (see below).
Note: The master switch is greyed out until you've added at least one additional language. Adding a language alone doesn't enable switching — you also need to flip the switch.
Auto-Detect: The 15 Supported Locales
When multi-language mode is on, your agent can auto-detect a caller's language only when both the primary and additional languages come from this set of 15 locales:
| Language | Locale(s) |
|---|---|
| English | en-US, en-GB, en-AU, en-CA, en-IN |
| Spanish | es-ES, es-MX |
| French | fr-FR, fr-CA |
| Chinese (Mandarin) | zh-CN |
| German | de-DE |
| Italian | it-IT |
| Japanese | ja-JP |
| Korean | ko-KR |
| Hindi | hi-IN |
If every supported language on your agent falls inside this set, auto-detection works seamlessly: the caller can switch languages mid-sentence and the agent follows.
Outside the 15: Single-Language Fallback
Languages outside the 15-locale auto-detect set are fully usable as your primary language or as a destination the AI switches to, but the agent cannot auto-detect them in the same way. Examples include:
- Thai, Vietnamese, Cantonese (zh-HK)
- Arabic, Dutch, Malay
- Filipino, Indonesian, Turkish, Tamil, and others
When the agent's current language is one of these out-of-set languages, the call temporarily runs in single-language mode. The agent can still switch into one of these languages if it understands the caller's request — for example, a caller saying "Can you speak Thai?" in English — but it relies on the words being recognisable in the current session's language first.
Tip: If your customer base mostly speaks one out-of-set language with occasional English speakers, set the out-of-set language as primary and add English as an additional language. The agent will start in your primary language and switch to English when needed.
Voice Quality Tiers: HD vs SD
Each voice in Sidebar → AI Frontdesk → Voice & Languages → Voice carries a tier badge:
- HD — Higher-quality, more natural-sounding voices. Recommended for customer-facing agents where voice quality matters most.
- SD — Standard-quality voices. Useful when you need a specific voice character that isn't available in HD.
Not every language has an HD voice in every accent. When you pick a voice for an additional language, the picker shows you only voices that match that language, with the tier badge on each card so you can compare at a glance.
What Happens When the Master Switch Is Off
Even if you've added five additional languages, turning the Enable additional languages switch off makes the agent monolingual again. Specifically:
- The agent will not attempt to detect or switch languages during a call
- The language-switching tool is not exposed to the AI
- Only the primary language voice and greeting are used
This is useful if you want to keep your additional language setup as a draft (voices, greetings, pronunciations all configured) but pause it temporarily — for example, while one of the voices is being re-recorded.
Per-Language Voices and Greetings
When the agent switches into an additional language, it also switches to:
- The voice you selected for that language
- The greeting translated for that language (if you've set one)
- The pronunciation rules specific to that language
If a particular language has no voice configured, the agent falls back to the primary language voice for that segment of the call. Configuring a dedicated voice per language produces the most natural result.
Common Questions
Can I add a language not in the picker? The picker shows every language Speako currently supports. If a language you need is missing, contact us — we add new languages as voice quality and STT coverage allow.
Why can't the agent auto-detect Thai? Thai is outside the 15-locale auto-detect set. The underlying real-time speech model only supports detection across those 15 locales. Thai is still fully supported as a primary language or as a switch destination.
Does the AI understand when a caller asks to switch languages? Yes. If you've enabled additional languages, the agent listens for direct or indirect requests to switch ("Can we speak in Spanish?", "Hablamos español?") and switches immediately.
What happens if my caller mixes two languages in one sentence? Inside the 15-locale auto-detect set, the agent follows the dominant language of the sentence. For mixed input that crosses the auto-detect boundary, the agent stays in its current language until a full utterance in another language triggers a switch.
Do I need to save my changes for them to apply to live calls? Yes. Language changes — primary language, additional languages, voices per language, and the master switch — only take effect on live calls after you click Save Changes on the Voice & Languages page. Speako publishes the updated configuration to your agent automatically once saved.