Languages

EdgeVox supports 15 languages with automatic STT/TTS backend selection.

Kokoro Languages (Near-Commercial Quality)

These languages have native TTS support via Kokoro-82M:

Language	Code	STT	TTS	Default Voice
English	en	Whisper	Kokoro (a)	af_heart
English (British)	en-gb	Whisper	Kokoro (b)	bf_emma
French	fr	Whisper	Kokoro (f)	ff_siwis
Spanish	es	Whisper	Kokoro (e)	ef_dora
Hindi	hi	Whisper	Kokoro (h)	hf_alpha
Italian	it	Whisper	Kokoro (i)	if_sara
Portuguese	pt	Whisper	Kokoro (p)	pf_dora
Japanese	ja	Whisper	Kokoro (j)	jf_alpha
Chinese	zh	Whisper	Kokoro (z)	zf_xiaobei

Vietnamese (Specialized)

Vietnamese uses dedicated models for best accuracy:

Component	Model	Details
STT	Sherpa-ONNX Zipformer (30M, int8)	RTF ~0.01 on CPU, Apache 2.0
TTS	Piper ONNX	3 voices: vi-vais1000, vi-25hours, vi-vivos

Falls back to Whisper STT if Sherpa-ONNX is unavailable.

Piper Languages (Lightweight ONNX)

These languages use Piper VITS models — lightweight and real-time on CPU:

Language	Code	STT	Voices
German	de	Whisper	10 voices (de-thorsten, de-kerstin, de-ramona, ...)
Russian	ru	Whisper	4 voices (ru-irina, ru-dmitri, ru-denis, ru-ruslan)
Arabic	ar	Whisper	2 voices (ar-kareem, ar-kareem-low)
Indonesian	id	Whisper	1 voice (id-news)

Korean (Supertonic)

Korean uses the Supertonic-2 ONNX model — real-time on CPU with 10 voice styles:

Voice	Description
ko-F1 .. ko-F5	5 female voices (calm, bright, clear, crisp, kind)
ko-M1 .. ko-M5	5 male voices (lively, deep, polished, soft, warm)

Thai (PyThaiTTS)

Thai uses PyThaiTTS with a Tacotron2 ONNX model (Apache 2.0):

Voice	Model	Sample Rate
th-default	lunarlist_onnx	22,050 Hz

Switching Languages

Via TUI

/lang fr          # Switch to French
/lang vi          # Switch to Vietnamese
/lang ko          # Switch to Korean (Supertonic TTS)
/langs            # List all languages with backends
/voices           # List voices for current language

Or use the Language and Voice dropdowns in the side panel.

Via Web UI

Use the language and voice selectors in the web interface, or type /lang ko in the text input.

Via CLI

edgevox --language fr
edgevox --language ko --voice ko-M2
edgevox --web-ui --language de --voice de-thorsten

Via Code

from edgevox.core.config import get_lang, LANGUAGES

cfg = get_lang("ja")
print(cfg.name)          # "Japanese"
print(cfg.stt_backend)   # "whisper"
print(cfg.tts_backend)   # "kokoro"
print(cfg.kokoro_lang)   # "j"
print(cfg.default_voice) # "jf_alpha"

cfg = get_lang("ko")
print(cfg.tts_backend)   # "supertonic"
print(cfg.default_voice) # "ko-F1"

Adding a New Language

Add one entry to edgevox/core/config.py:

_reg(LanguageConfig(
    code="tr",
    name="Turkish",
    kokoro_lang="a",          # fallback to English TTS
    default_voice="af_heart",
    test_phrase="Merhaba.",
))

The language will automatically appear in:

• TUI language and voice dropdowns
• Web UI selectors
• /langs command output
• CLI --language option