Hi Product Hunt, we are Paxa Labs, an AI research lab in Bangkok. Thai breaks the assumptions most language tooling makes. Words are written without spaces, tone changes the word, speakers switch between Thai and English inside one sentence, and business documents stack both scripts with tables, stamps, and handwriting on a single page. General-purpose models treat all of that as edge cases. For anyone building for Thai users, it is the input. We build one focused model around each place a general pipeline breaks, and ship them as one API: - Text to speech: Paxa Flash, our proprietary Thai speech model. 26 voices including Isan, Northern, and Southern accents, MP3/Opus/WAV, streaming, word and sentence timestamps, a live WebSocket endpoint, and an OpenAI-compatible alias so an existing client only needs a base URL change. - OCR: our proprietary document model for PDF and images, returning per-page Markdown or typed blocks (headings, tables, lists, figures). - Translation: 14 source languages into Thai, with formality, glossary, do-not-translate, and Markdown/HTML modes that leave the markup where it is. Everything bills in credits, 1,000 to a dollar: $15 per million characters for speech, $25 per million for translation, $0.0065 per page for OCR. Every account starts with free credits, every request is charged before inference and refunded if it fails, and the playground lets you try all three in the browser before writing code. The docs are also served as Markdown with an llms.txt index and an installable agent skill (npx skills add https://paxalabs.com), so your coding agent can read the contract directly. Speech recognition is still in research. It ships when it survives real Thai audio. If you build for Thai users and something a general model gets wrong is costing you, tell us here or at paxalabs.com/contact. We read everything.

Paxa Labs

Thai-native text to speech, OCR, and translation, one API

Thai-native text to speech, OCR, and translation, one API