Core settings for onex-voiceInteraction are defined in shared/config.lua.
| Type | string |
| Default | "en" |
Sets the language for speech recognition. Use an ISO 639-1 code (e.g. "en", "fr", "de") or a regional variant (e.g. "pt-BR", "zh-CN", "es-MX") to transcribe in that language, or "auto" to detect the player's spoken language automatically.
"auto" requires BYOK (your own API key) — it detects any spoken language and converts it to English before matching, so your English voice command setup works for all players. Without BYOK, "auto" falls back to "en".
| Type | string | nil |
| Default | nil |
Sets the language for NPC text-to-speech responses. Leave as nil to match Config.Language.
| Type | string |
| Default | "native" |
| Values | "romanized" | "native" | "original" |
Controls which script is displayed in NPC speech subtitles when ResponseLanguage is a non-Latin language.
"native" — show native script (e.g. "آپ کیسے ہیں" for Urdu, "आप कैसे हैं" for Hindi)"romanized" — show romanized text matching TTS input (e.g. "aap kaise hain")"original" — show original text before translation (e.g. English source text)Has no effect when the response language uses a Latin script, or when no translation is active.
| Type | boolean |
| Default | true |
When ResponseLanguage is a non-Latin language, overrides the NPC's configured voice with a language-native Edge TTS model and sends native script text to the synthesizer. For example, ResponseLanguage = "hi" uses hi-IN-MadhurNeural (male) or hi-IN-SwaraNeural (female). Falls back to the NPC's configured voice if no native model is defined for the language.
Has no effect when using ElevenLabs or Azure TTS. Set to false to always use the NPC's configured voice with romanized text instead.
| Type | boolean |
| Default | true |
Converts numbers to words before translation (e.g. "2 dollars" → "two dollars"). Improves regional language TTS by letting the translation engine handle number words natively. Only applies when translation is active (ResponseLanguage is set or Language is not English).
Translation is automatic when Config.Language differs from the language your command phrases are written in. Advanced translation behaviour is controlled via server convars in server.cfg — not in config.lua:
| Convar | Default | Description |
|---|---|---|
voice_translate | "auto" | "auto" | "true" | "false" — enable/disable translation |
voice_translate_target | "en" | Language your command phrases are written in |
voice_dual_match | "false" | Match against both raw and translated text in parallel |
voice_dual_match_priority | "highest_score" | "highest_score" | "raw_first" | "translated_first" |
# server.cfg example
set voice_translate "auto"
set voice_translate_target "en"
set voice_dual_match "true"
set voice_dual_match_priority "highest_score"
| Type | string |
| Default | 'G' |
The key players hold to activate voice recognition. Must be a valid key name from Config.AvailableVoiceActivationKeys. Players can rebind this from the F2 settings panel.
| Type | string |
| Default | 'X' |
The key players press to cancel NPC speech while it is playing.
Controls the F2 voice settings panel.
| Type | boolean |
| Default | false |
Prints detailed voice recognition and matching output to the F8 client console. Can also be enabled at runtime via the onex_voice_debug convar. Recommended to keep false in production.
| Type | boolean |
| Default | false |
Enables additional debug output from all addons (shops, ambient NPC, banking, etc.). Can also be enabled via the onex_voice_addondebug convar. Set false in production.
| Type | boolean |
| Default | true |
Displays the recognized speech as text beneath the audio spectrum in the NUI. Color-coded by state: white while listening, yellow while processing, green on match, red on no-match.
| Type | boolean |
| Default | false |
Enables a hidden voice command — say "voice recognition test" — that returns a confirmation response. Useful for verifying STT is working without triggering any addon actions.
| Type | boolean |
| Default | true |
Master toggle for broadcasting NPC TTS speech to nearby players as positional audio. Each addon also has its own networkSync toggle — both must be true for a given addon's audio to broadcast.
Settings for LLM-based matching and addon AI. Only applies with a BYOK LLM API key configured.
Voice assignment mode when ElevenLabs TTS is active (BYOK only).
| Type | boolean |
| Default | false |
Enables a custom speech correction dictionary sent to the STT engine. Accepts up to 200 entries, each pattern and replacement up to 50 characters.
Pattern → replacement mappings applied before intent matching. Patterns are case-insensitive and word-boundary enforced.
| Type | table |
Sends vocabulary hint lists to the STT engine to improve recognition of domain-specific terms.
All hint terms are filtered against registered command phrases before being sent to the backend — only words that appear in at least one registered phrase are included. Per-addon hints are defined in each addon's promptHints field in Config.Addons.
Controls the in-world voice hint display — floating phrase bubbles above NPCs and the 2D screen overlay.
Each addon can be individually enabled or disabled. The full defaults from shared/config.lua:
Set an addon's enabled to false to fully disable it. Disabled addons do not load their config, proximity zones, or prompt hints. The networkSync field (not networkAudio) controls whether NPC speech is broadcast to nearby players.
Last updated 3 months ago