Rivo Voice

Modern voice command system for FiveM. Players hold a key and speak, We transcribes their audio, and your script reacts.

Free

  • Standard: Free
  • Open Source: $50.00

By Rivo

Rivo Voice

Description

Stratis Voice

Real voice commands for FiveM. Speak naturally, your scripts react in 300 milliseconds.
Built for police roleplay, civilian RP, shops, mechanics, dispatch radio, anywhere typing /me or pressing context menu buttons slows the moment down.

The problem with every other voice resource

  • Browser webkitSpeechRecognition was killed in FiveM CEF years ago. Looks like it works, returns garbage.
  • Older free transcription APIs are so inaccurate that "engine on" gets transcribed as "energy hat" half the time. Players give up after the third misfire.
  • Push to talk on the same key as voice chat creates constant key conflicts. Players stop using it.

How Stratis Voice is different

A modern AI speech recognition model running at over 200x realtime. Accurate even with background music, accents, and 50+ languages. Two separate keybinds so voice mode never fights your voice chat key. A live waveform UI that pulses with your voice. Hard cost caps and replay protection built in from day one.

How it feels in use

1. Press F7 once. The mic icon appears with a soft green glow.
2. Hold N and say "hands up".
3. Release N.
4. About 300ms later your character throws their hands up.
5. Press F7 again. Mic disappears, talk key does nothing.

No spammy GTA notifications. No menu nag. The HUD itself shows everything: idle, armed, recording, processing.

Features

Voice capture

  • Adaptive VAD learns your background noise level on the fly
  • Pre-roll buffer so the first syllable of your sentence is never clipped
  • Live waveform visualization in the HUD reacts to your voice in real time
  • Microphone picker UI lists every input device and remembers your choice across sessions
  • Two key system: F7 toggles voice mode, N records, both fully rebindable
  • Wake word filter for servers that want extra protection against accidents

Recognition quality

  • Custom phrase biasing teaches the AI your server's vocabulary
  • Police 10-codes (10-4, 10-99, Signal 100) work out of the box
  • NATO phonetic alphabet, unit callsigns like 1-Adam-10 or 2-Boy-15
  • Gang names, shop names, locations, anything custom
  • Hallucination rejection server side so empty or unclear audio never fires false actions
  • Confidence scoring on every transcription using internal probability data
  • 50+ language support with auto-detect or pinned language

Production safety

  • Daily transcription caps per player AND server wide
  • Per-player rate limit with a minimum gap between requests
  • Audio fingerprint deduplication catches replay attacks and double submits
  • Concurrency cap keeps 100 players hammering at once from blowing past API limits
  • State machine watchdog auto-recovers from any stall
  • Persistent quota tracking survives server restarts
  • Blacklist policy can drop or report players for configured terms
  • Token never reaches the client, even if your script is leaked

Action system

  • Three action types: client event, server event, console command
  • Custom Lua function hook for advanced actions
  • Runtime exports to add or remove commands dynamically without restarting
  • Reverse word matching so phrase order does not matter
  • Block in vehicle flag for actions that should not fire while driving
  • Per-action confidence threshold for fine grained control

Real use cases

Police roleplay
"1-Adam-10 to dispatch, 10-4, en route to Mission Row, requesting Code 3 backup."

Civilian RP
"Hands up." "Engine off." "Open the trunk." "Lock the doors."

Shops and commerce
"Buy this." "Checkout." "Deposit five thousand."

Mechanic
"Engine on." "Check tires." "Headlights."

Dispatch radio
"Show 1-Adam-10 10-23 at Davis. Code 4."

Performance

  • First word to action: about 300ms. Most competitors sit between 1 and 2 seconds.
  • Accuracy on roleplay slang is high with prompt biasing. Competitors using older APIs are a coin flip.
  • 50+ languages supported with auto-detect. Most competitors are English only.
  • 5 layered defenses against spam including rate limit, fingerprint dedup, daily caps, and concurrency control. Competitors typically have none.
  • Tested at 100 simultaneous players without dropping requests.

Pricing math

Real world cost for a 32 player server running 4 hours a day for 30 days is about $5 a month. Free tier covers 2000 requests a day before you pay anything. Daily caps prevent any worst case from going past your budget.

What is in the box

  • Drop in resource folder
  • README with full setup and config documentation
  • Customizable example actions (hands up, engine on, engine off)
  • Every keybind, prompt, rate limit, quota exposed in shared/config.lua
  • Microphone picker UI styled to match the HUD
  • Two minutes from download to working voice commands

Setup

  • Download the resource
  • Get an API token (free signup, no credit card)
  • Paste it into server/config.lua
  • Restart

That's it. Working voice commands.

Tags

  • paid
  • standalone
  • voice script

Rivo Voice documentation