One platform connecting startups, corporations, research institutions, regions and investors
Playbook & Handbook · February 6, 2026
Winning Accelerator Playbook 2026 by BambuUP
EDTECH: LEVERAGING AI-POWERED TECHNOLOGIES TO TRANSFORM THE LEARNING EXPERIENCE
1. Technology overview
Paxa Labs describes itself as a research lab in Bangkok building speech and language AI for English, Thai and Mandarin. It targets the failure points of general-purpose models on Thai, such as missing word spaces, tones and Thai-English code-switching.
The platform offers text-to-speech (Paxa TTS Flash v2 with 50 voices), speech-to-text in research preview, OCR and extraction for business documents, and AI translation. Developers can use streaming, word and sentence timestamps, a WebSocket endpoint and an OpenAI-compatible speech endpoint. The company cites 15-20 ms to first audio in its own load test on dedicated hardware.
2. Highlighted features & specifications
- Text-to-speech: Paxa TTS Flash v2 with 50 voices across English, Thai and Mandarin.
- Speech-to-text: research-preview transcription of Thai, including turn-by-turn real-time mode.
- OCR and extraction: Thai business documents such as invoices and contracts.
- AI translation: translation with glossary and formality controls.
- Developer API: streaming, timestamps, WebSocket and an OpenAI-compatible speech endpoint.
- Voice design: custom voice creation feature announced in October 2026.
3. Potential application
Thai-language voice announcements and IVR for Vietnamese manufacturers with Thai customers. OCR of Thai invoices and customs documents for logistics firms. Thai audio versions of product content for e-commerce.
4. Benefits
Pricing is published: TTS v2 at $10 per 1M characters, OCR at about 6.5 credits per page, and new users get 100 free credits with no card. Failed requests are refunded. An OpenAI-compatible speech endpoint eases integration into existing voice-agent stacks.
5. Market trends & opportunities
Thai call centres, audiobook producers and document-heavy back offices need speech and OCR models that handle Thai script and tonal speech better than global models. Small specialised vendors are emerging alongside large Thai institutions.
6. Other information
Based in Bangkok per the website. Blog posts dated 2026-09 and 2026-10 cover STT Lite Realtime, TTS Flash v2 (2026-10-16) and voice design. Mandarin support was added in 2026-09. No customer list, certifications or founding year are stated on the pages fetched.
