nrl-ai/edgevox-models

EdgeVox Models

Consolidated STT and TTS models for EdgeVox — sub-second local voice AI for robots and edge devices.

STT Models

Model	License	Original Source
sherpa-zipformer-vi-30M-int8	Apache-2.0	csukuangfj2/sherpa-onnx-zipformer-vi-30M-int8-2026-02-09

Piper Voices

All Piper voices are ONNX models based on the VITS architecture. They run in real-time on CPU.

Voice	License	Unrestricted?	Original Source
vi-vais1000	CC-BY-4.0	Y	speaches-ai/piper-vi_VN-vais1000-medium
vi-25hours	Unknown	?	speaches-ai/piper-vi_VN-25hours_single-low
vi-vivos	CC-BY-NC-SA-4.0	?	speaches-ai/piper-vi_VN-vivos-x_low
de-thorsten-high	CC0-1.0	Y	speaches-ai/piper-de_DE-thorsten-high
de-thorsten	CC0-1.0	Y	speaches-ai/piper-de_DE-thorsten-medium
de-thorsten-low	CC0-1.0	Y	speaches-ai/piper-de_DE-thorsten-low
de-thorsten-emotional	CC0-1.0	Y	speaches-ai/piper-de_DE-thorsten_emotional-medium
de-kerstin	CC0-1.0	Y	speaches-ai/piper-de_DE-kerstin-low
de-ramona	M-AILABS (unverifiable)	?	speaches-ai/piper-de_DE-ramona-low
de-eva	M-AILABS (unverifiable)	?	speaches-ai/piper-de_DE-eva_k-x_low
de-karlsson	M-AILABS (unverifiable)	?	speaches-ai/piper-de_DE-karlsson-low
de-pavoque	CC-BY-NC-SA-4.0	?	speaches-ai/piper-de_DE-pavoque-low
de-mls	CC-BY-4.0	Y	speaches-ai/piper-de_DE-mls-medium
ru-irina	Unknown	?	speaches-ai/piper-ru_RU-irina-medium
ru-dmitri	CC0-1.0	Y	speaches-ai/piper-ru_RU-dmitri-medium
ru-denis	CC0-1.0	Y	speaches-ai/piper-ru_RU-denis-medium
ru-ruslan	CC-BY-NC-SA-4.0	?	speaches-ai/piper-ru_RU-ruslan-medium
ar-kareem	No license	?	speaches-ai/piper-ar_JO-kareem-medium
ar-kareem-low	No license	?	speaches-ai/piper-ar_JO-kareem-low
id-news	No license	?	giganticlab/piper-id_ID-news_tts-medium

Legend: Y = freely redistributable, ? = license unclear or has restrictions (see per-voice README)

Other TTS Models (also included in this repo)

Model	Languages	License	Size
Kokoro-82M	en, en-gb, fr, es, hi, it, pt, ja, zh (25 voices)	Apache 2.0	~338 MB
Supertonic-2	ko, en, es, pt, fr (10 voices)	MIT (code) + OpenRAIL-M (weights)	~255 MB
PyThaiTTS	th (1 voice)	Apache 2.0	~163 MB

Usage

from huggingface_hub import hf_hub_download

# Piper voice
model = hf_hub_download("nrl-ai/edgevox-models", "piper/de-thorsten/model.onnx")
config = hf_hub_download("nrl-ai/edgevox-models", "piper/de-thorsten/config.json")

# Kokoro
kokoro_model = hf_hub_download("nrl-ai/edgevox-models", "kokoro/kokoro-v1.0.onnx")
kokoro_voices = hf_hub_download("nrl-ai/edgevox-models", "kokoro/voices-v1.0.bin")

# Supertonic
sup_encoder = hf_hub_download("nrl-ai/edgevox-models", "supertonic/onnx/text_encoder.onnx")

License

This repository contains models under mixed licenses. Each voice subdirectory has its own README with license details. See individual voice directories for specifics.

Models marked as CC0 or CC-BY-4.0 are freely usable. Models with unknown or NC/SA licenses are included for convenience with clear notes — check the license before commercial use.