Nia is a cascaded, on-device voice stack — Whisper ears, a Gemma brain, a Kokoro voice — wired together so every part is a swappable box. No cloud, no per-minute meter, no data leaving the machine.
Every stage of the pipeline is exposed as a tool. Talk, interpret, dub, synthesize, clone and transcribe — all from the same on-device engine.
Talk to Nia in real time. Tune voice, tone, speed and language live — no reconnect.
Two-way live interpretation. Speak one language, hear another, turn by turn.
Redub audio into a new language while preserving the speaker's voice and timing.
Turn text into speech across voices and languages. Preview, then download WAV or MP3.
Clone a voice from a short reference and give Nia a consistent character.
Drop in audio and get accurate transcripts with MLX Whisper, on your Mac.
A cascaded pipeline: audio in, audio out, every box swappable behind a stable interface. Change the brain or the voice without rewiring.
Type once, add voices and languages, and generate them all in a batch. Preview each clip inline, then download WAV or MP3.
Open Voices StudioInterpret a conversation turn by turn, or redub a whole clip into a new language while preserving the original speaker's voice and timing.
Try the studiosA simple, OpenAI-shaped API for speech, voices and chat. Mint a key, call an endpoint, and metering shows up in your dashboard.
curl https://localhost/v1/speech \
-H "Authorization: Bearer $NIA_KEY" \
-H "Content-Type: application/json" \
-d '{
"voice": "af_heart",
"language": "en",
"input": "Hello from Nia."
}' --output reply.wavNo sign-up, no cloud round-trip. Grant mic access and start the conversation.
Open the Playground