Supported Languages

FinnAI supports 46 languages out of the box. This page lists every supported language, the quality tier for each, and how to set up multilingual Finns.

Quick reference. Most users only need English or one regional language. Pick what your callers actually speak — not what your website is in. If you're not sure, look at where your customers are physically located.

Full language list

The Finn configuration wizard ("Create New Finn → Language") shows all 46 supported languages with their flags. Here they are:

Quality tiers

Not all languages have the same quality. We grade them in three tiers based on voice naturalness, STT accuracy, and regional accent coverage.

Tier 1 — Production-grade

Indistinguishable from a human at typical phone audio quality. Multiple voice options per language including regional variants. STT handles accents, code-switching, and ambient noise well.

• English (US, UK, Australian, Indian, Canadian)
• Spanish (Spain, Latin American, Mexican)
• French (France, Canadian)
• German
• Portuguese (Brazil, Portugal)
• Italian
• Japanese
• Mandarin Chinese
• Hindi
• Arabic (MSA + Egyptian)
• Korean

Tier 2 — Good with caveats

Natural-sounding voice but limited regional variants. STT works well in quiet environments but may struggle with heavy accents.

• Bengali
• Urdu
• Russian
• Dutch
• Polish
• Turkish
• Swedish, Norwegian, Danish
• Vietnamese, Thai, Indonesian
• Tamil, Telugu, Marathi, Gujarati

Tier 3 — Functional

Voice is clearly synthetic but understandable. STT requires good audio quality. Best for low-stakes use cases.

• Kannada, Malayalam, Punjabi
• Filipino, Malay
• Hebrew, Greek
• Finnish, Hungarian, Czech, Romanian
• Bulgarian, Croatian, Serbian, Slovak, Slovenian, Ukrainian
• Estonian, Latvian, Lithuanian

If your use case is high-stakes (sales, healthcare, finance), prefer Tier 1 languages. For internal tools, surveys, or notifications, Tier 2/3 are fine.

Setting language on a Finn

When you create or edit a Finn:

1. Identity tab → Language dropdown.
2. Pick the primary language.
3. The voice library filters to voices that support that language. Pick one.
4. Save.

The Finn will speak the chosen language for the entire call. STT will be tuned to recognize speech in that language.

Multilingual Finns

Some use cases need a Finn that can switch languages mid-call (e.g. a receptionist for a clinic serving both English and Spanish speakers).

Option A — Language detection at call start

The Finn opens in a primary language, then auto-detects the caller's language from their first response and switches.

To enable: Identity tab → Multilingual mode → Auto-detect. Select up to 3 candidate languages.

Pros: seamless caller experience. Cons: adds ~1 second of latency on the first switch. Only works between Tier 1 languages reliably.

Option B — Explicit language menu

Finn opens with: "For English, say English. Para español, diga español." Caller's choice routes them into the appropriate flow.

To enable: configure a Workflow with a Decision node based on detected language at the first Ask node.

Pros: explicit and reliable. Cons: feels more like an IVR. Best for jurisdictions where multilingual disclosure is required.

Option C — Separate Finns per language

Run two completely separate Finns. Route inbound based on the dialed DID (one number per language) or based on a caller property in your CRM.

Pros: independently optimized per language. Cons: double the maintenance.

Most accounts use Option A. Option C makes sense when each language audience is large enough to warrant a dedicated team.

Voice + language pairing

Not every voice supports every language. The voice library shows supported languages per voice.

Common patterns:

• English speakers — broadest voice selection (50+ voices across tiers, accents, genders).
• Spanish, French, German — 15-25 voices each with regional variants.
• Indian languages — 3-8 voices per language, with cross-tier mix.
• Tier 3 languages — typically 1-3 voices each.

When in doubt: pick the language first, then the voice from filtered options.

Custom vocabulary per language

The custom vocabulary field (under Identity → Custom Vocabulary) accepts words in any language. Use it to teach the Finn:

• Brand names (especially non-English brand names with unusual spellings)
• Product SKUs
• Industry-specific terminology
• Common caller name pronunciations

Example for an English Finn serving Indian customers: ["Ola", "Zomato", "Razorpay", "Aadhaar", "PAN card", "GST"]

Custom vocabulary works across languages — particularly useful when callers code-switch (mix two languages in one sentence).

Language and the knowledge base

The knowledge base is language-agnostic. You can upload documents in any language. At call time, the Finn retrieves relevant content regardless of the source document's language and translates as needed in its response.

Practical advice: upload knowledge in the language your Finn primarily speaks. Cross-language retrieval works but is less reliable.

Code-switching (mixing languages mid-sentence)

In some regions, callers naturally mix languages — Hindi-English ("Hinglish"), Spanish-English ("Spanglish"), French-Arabic (in North Africa). Tier 1 voices handle this well; the Finn will follow the caller's lead.

To explicitly enable: Identity → Style & Guardrails → add:

Match the caller's speaking style — if they mix English and Hindi, you can mix too. Don't correct or comment on their language choice.

This makes the Finn feel natural to bilingual callers.

Languages we don't support yet

If your language isn't on the list above, we don't support it natively yet. Options:

• Use the closest related language. Some adjacent languages work surprisingly well (e.g. Catalan callers often comfortable with Spanish Finn).
• Request the language — email support@hirefinn.ai with the language and expected volume. We prioritize based on demand.
• Enterprise custom builds — Enterprise plans can request custom voice + STT models for specific languages or dialects, typically with a one-time setup fee.

Common questions

Q: Can my Finn handle accents — like an English-speaking Finn talking to a strong Glaswegian or Indian-English speaker? A: Tier 1 English is robust to most accents. Heavy regional accents may have higher STT error rates. Test with a few representative callers.

Q: What's the latency difference between languages? A: Tier 1 languages: <800ms. Tier 2: 800-1000ms. Tier 3: 1000-1300ms. Above ~1.5s feels unnatural.

Q: Can the Finn translate between languages mid-call? A: Not as a primary function — that's a different product (real-time translation). The Finn speaks one language at a time. Multilingual mode (Option A above) is the closest equivalent.

Q: Do I pay more for non-English languages? A: No. All languages cost the same per minute.

Next

• Creating a Finn → Identity & voice — set up language and voice.
• Voice & Personality Gallery → — templates by use case.
• Compliance → — country-specific rules that vary by region.