Introduction

hama is a cross-platform G2P and phoneme ASR runtime for Python and TypeScript.

Guiding principles

• On-device. Inference runs locally in Python, Node/Bun, and the browser.

• Plug-and-play. Packaged assets and consistent APIs make hama easy to drop in and run across environments.

• Explainable. G2P returns alignments, and ASR returns collapsed and frame-level token views.

Supported languages

hama ships vocabularies for English, Korean, Chinese (simplified/traditional), Dutch, French, German, Italian, Japanese, Portuguese, Spanish, and Thai. Adding a language means regenerating the vocab and ONNX assets; see Architecture and Modalities for the workflow.

For the shortest path, start with Quickstart, Installation, and APIs.