Hear Nia speak
Search
Live voice
API keys
Open appNative voices, one tap to replay.
Multi-speaker English and European voices, a native neural voice for each Indian language, and twelve Arabic dialects.
Talk to it. Interrupt it. It keeps up.
Two-way calls with natural turn-taking and barge-in. Run Nia's own brain, or plug in any OpenAI-compatible LLM and keep your knowledge on your side.
Each Indian language gets its own native neural voice, so Tamil sounds Tamil and Punjabi sounds Punjabi.
Nia handles the ears and the voice; your LLM and RAG write every word, over the open call.
Personas bundle a base voice with a tone, or clone one from a short clip, through /v1/personas.
0.6sWebSocket speech at 24 kHz and live transcripts with interim text, with no TURN server needed.
One key. Every voice job.
Text to natural speech in 33 languages. WAV or MP3.
Speech to text with script-aware language ID for Indic languages.
The text engine behind the live Interpreter. Auto-detects the source.
Transcribe, translate and re-speak a clip, optionally in the original voice.
Built for real conversations.
Take live calls, interpret across languages and dub your content, in the language the person actually speaks.
Ship a voice in two commands.
Bearer-token auth on console.niavoices.com. Keep your key in NIA_KEY, never in a browser. Scoped keys, one base URL.
curl -s https://console.niavoices.com/v1/speech \ -H "Authorization: Bearer $NIA_KEY" \ -H "Content-Type: application/json" \ -d '{"text":"Hello from the Nia API.", "voice":"nia-en-us-1"}' \ --output hello.wav # POST /v1/speech · speech scope
33
Languages on /v1/speech0.6s
English speech returned2.6s
Hindi transcription of a 9s clip12
Arabic dialectsOur models. Your data. Your brain.
Every endpoint runs on Nia's own engine, not a third-party API behind a wrapper.
Live audio is peer-to-peer. Usage records keep durations and token counts, never what was said.
Nia handles the ears and the voice; your RAG and LLM write every word, over the open call.
Ears, brain and voice.
Questions, answered.
33 on /v1/speech: Hindi, Tamil, Malayalam, Telugu, Kannada, Bengali, Marathi, Gujarati, Punjabi, Odia, Assamese, Urdu and Maithili, plus English, Spanish and twelve Arabic dialects.
No — Nia ships a default brain. Bring your own OpenAI-compatible endpoint any time to use your own knowledge and tone instead.
No. Live audio is peer-to-peer and never written to disk; usage records keep only durations and token counts.
Access is invite-only while we scale — request an invite and you're live in minutes once it's approved.
Yes, from a 3–40 second clip through /v1/personas, or pick one of the native neural voices bundled with a tone.
Give your product a voice from India.
Access is invite-only while we scale. Ask for a code and you're live in minutes.