Paxa Labs Launches Thai TTS Model with 32 Voices and Open Research Weights

A Bangkok-based AI research team called Paxa Labs has released Paxa TTS Flash, a text-to-speech model supporting Thai, English, and code-switching within a single sentence, offered via a commercial API with 32 distinct voices. Alongside it, the team published a separate research artifact called Wayu-Paxa-TTS-Edge, an 82-million-parameter model freely downloadable under CC BY-NC 4.0 and runnable on a CPU, intended for reproducibility rather than production use. The project involves Kunat Pipatanakul, a co-creator of the Typhoon language model, and was developed in collaboration with Wayu Research, a non-profit AI lab in Bangkok. The team built the system specifically to handle structural challenges unique to Thai, including unspaced words, tonal variation, and mixed-language business documents. A minor discrepancy exists between the launch announcement citing 26 voices and the current product page listing 32, though the live voice listing on the website confirms the higher figure.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in