All capabilities

Models & languages

Your models. Your languages. Swappable per agent.

Choose the STT, TTS and LLM for each agent, and sound truly native to any regional speaker — with per-language pronunciation packs, not a one-size-fits-all voice.

Configure per agent

Models & add-ons

Voice pipeline · per agent

Editable

STT

SonioxRealtime

$0.005

LLM

Google Gemini2.5 Flash

$0.020

TTS

Smallest AILightning v3.1

$0.035
Estimated cost / min$0.06

Model stack for this agent

Deepgram Nova-2
Gemini 2.5 Flash
ElevenLabs Flash v2.5

Swap any model per agent — no rebuild.

Languages

EnglishSpanishFrenchGermanPortugueseItalianMandarinJapaneseArabicHindiKorean+ 66 more

What you get

  • Pick STT (Deepgram, Soniox, AssemblyAI, ElevenLabs, Azure, Sarvam) per agent.
  • Pick TTS (ElevenLabs Flash/Turbo/V3, Cartesia, Smallest AI, Azure, Gemini, Sarvam).
  • Pick your LLM (Gemini 2.5 Flash and others) or a realtime model like Gemini Live.
  • Native pronunciation & normalization packs per language — Arabic, Chinese (zh/ja/ko), Spanish, Portuguese and per-language Indian packs — so numbers, names and nuances sound right, never transliterated.
  • Evidence-based runtime language switching: the agent detects the caller's language mid-call and switches voice and prompting — handling code-switching and caller-requested changes.
  • Locale-aware number, currency and script normalization (e.g. Indian digit grouping, lakh/crore) so figures are spoken naturally.
  • 77+ languages across Indic, Arabic, CJK, European and Southeast-Asian families — real telephonic audio with noise handling and echo cancellation.
  • Swap any model any time — no rebuild, no lock-in.

Sound local everywhere — the same platform is native to a caller in Riyadh, Mumbai, São Paulo or Shanghai, not just English-first with an accent.

Run this on a real call.

Book a demo and see it live, end-to-end.

Book a demo