Kiri is a voice platform built in Phnom Penh. We make text to speech, transcription and voice cloning for a language most speech tools were never trained to handle.
Why we started
Khmer writes without spaces between words, stacks its consonants, and has far less public training data than English. Speech tools built elsewhere treat it as an afterthought, when they support it at all. Cambodians heard robotic voices and read transcripts full of guesses.
Kiri exists to close that gap. We train and tune our own models for Khmer, and we measure them on Khmer, not on an average across a hundred languages where ours quietly loses.
English and other languages work too. Khmer is simply the one we refuse to get wrong.
What we build
Natural Khmer and English speech from written text, with emotional range and control over pace and delivery.
Transcripts with speaker labels, word-level timestamps and subtitle export in SRT, VTT, JSON, TXT and CSV.
A speaker’s own voice from a short reference recording, reusable across every generation.
A new voice described in words — tone, gender, age, pace and mood — previewed before you keep it.
A REST API with an OpenAI-compatible transcription endpoint, so existing code needs little more than a new base URL.
Projects, history, usage analytics and team billing for the people who work with audio every day.
Our parent company
Blizzer is an AI company in Phnom Penh building for Cambodia and the region. Kiri is where that work turns into voice.
Visit blizzer.tech