v1.2.2

What's changed

Streamlined Property Inspector

• Simplified setup: the inspector now only shows Microphone, Model and Commands.
• Removed the Python path, console port, corrections, vocabulary, minimum confidence, cooldown and debug settings. They now use their defaults, and the Full version needs no Python setup.

Key Icons

• Keys now use the radio icon SVGs for idle, thinking, sent, no game and no mic states instead of plain text.
• Icons are centred and no longer overlap the labels.
• The sent key shows the command that was sent in its second label.
• Fixed the thinking key's hard-to-read black text.

Packaging

• Fixed the manifest key image path, which was breaking streamdeck pack and the release build.

Speech Recognition & Commands

• Transcription confidence scoring, with improved speech normalisation (corrections, spoken numbers, NATO phonetic alphabet).
• Added plate and vehicle argument types, required arguments, minimum and maximum lengths, and per-command cooldowns.
• Matching prioritises the longest trigger phrase; callsign and registration validation is stricter.
• The default QVR command uses plate recognition rather than callsign formatting.
• Existing custom commands remain supported.
• Clearer feedback for rejected commands, low-confidence transcriptions and cooldowns.

---

Hold a Stream Deck key, say a command, and release. Your speech is transcribed locally on your PC and the resulting command is sent to the FiveM client console. Nothing is sent to the cloud.

Release

Which file do I download?

File	Pick this if
fivem-voicecontrols-full.streamDeckPlugin	You don't have Python. Includes its own speech engine (about 110 MB).
fivem-voicecontrols-lite.streamDeckPlugin	You already have Python 3.9+. Smaller download, but you need to install the packages yourself: pip install faster-whisper numpy sounddevice

Install

1. Download one of the files above and double-click it. Stream Deck will install the plugin.
2. Drag Voice Command from the FiveM-VoiceControls category onto a key.
3. Hold the key, speak, and release.
4. The first press downloads the speech model (about 150 MB). The key displays DOWNLOADING while this happens.

Features

• Push-to-talk speech recognition running entirely on your PC.
• Custom commands: define your own trigger phrases and FiveM console commands using JSON.
• Support for NATO phonetic alphabet, spoken numbers, and common mishearings.
• Speech corrections and custom vocabulary.
• Configurable argument types, validation, required arguments, and cooldowns.
• Configurable microphone selection.
• Transcription confidence threshold.
• Detailed parsing logs for troubleshooting.
• Key status indicators: LISTENING, THINKING, SENT, NO GAME, NOT HEARD, and error states.
• Audio is not saved unless Debug is enabled.

Known issues

• Windows only for now. The Full version is built for Windows. macOS is currently untested.
• Windows or antivirus warnings (Full version). The bundled engine is not code-signed, so Windows may show an unknown publisher warning. Select More info, then Run anyway. Some antivirus tools may also remove it. If this happens, try the Lite version.
• FiveM must be running with its console port available (default 29200), or the key will display NO GAME.

Troubleshooting

Key shows	Meaning
NO PYTHON	Lite only. Python wasn't found. Set its full path in the key's Python setting.
PIP INSTALL	Lite only. Required Python packages are missing. Run the pip install command above.
MODEL FAIL	The model download failed. Check your internet connection.
NO MIC	The microphone couldn't be opened. Check the Microphone setting and your OS microphone permissions.
NO GAME	FiveM isn't running, or the console port doesn't match.
NOT HEARD	Nothing matched a command. Enable Debug to see what was recognised.
LOW CONF	The transcription confidence was below your configured minimum.
COOLDOWN	The command was ignored because its cooldown period has not expired.
BAD CONFIG	The command configuration is invalid. Check your JSON settings.

Full documentation is available in the README.

This is an unofficial community project and is not affiliated with FiveM, Cfx.re, or Elgato.

Antivirus note

If your antivirus removes listener.exe, download listener-windows.zip from this release and extract it into the plugin's bin\listener folder, or add an exclusion for the file.

Full Changelog: v1.2.1...v1.2.2

v1.2.1

What's changed

Enhanced Speech Recognition & Customisation

• Added speech corrections and custom vocabulary settings in the UI to improve recognition accuracy and handle common misinterpretations.
• Implemented transcription confidence scoring in listener.py, allowing commands below a configurable confidence threshold to be rejected automatically.
• Improved speech normalisation with separate processing stages for corrections, spoken numbers, and NATO phonetic alphabet conversion.
• Added detailed parsing traces to help identify how individual words are interpreted.

Expanded Voice Command Options

• Added support for additional argument types, including plate and vehicle, with improved validation.
• Added required argument settings and configurable minimum and maximum argument lengths.
• Introduced per-command cooldowns to prevent repeated commands from being triggered too frequently.
• Improved command matching to prioritise the longest matching trigger phrase, allowing more complex phrases and aliases.
• Improved callsign and registration validation to reject invalid or unreadable input.
• Updated the default QVR command to use plate recognition rather than callsign formatting.
• Improved handling of optional arguments, allowing commands to be triggered without additional information unless configured otherwise.

Developer Experience & Documentation

• Introduced detailed parsing trace logs to make debugging easier and provide visibility into how commands are interpreted.
• Updated README.md with explanations, examples, and configuration details for the new command options.
• Added formal release notes and documented the updated release workflow.

Compatibility & Behaviour

• Existing custom commands remain supported, although callsign arguments now use stricter validation.
• Improved handling of spoken letters and numbers, including retaining valid letters such as "A" when parsing registrations.
• Added clearer feedback for rejected commands, low-confidence transcriptions, and cooldowns.

---

Hold a Stream Deck key, say a command, and release. Your speech is transcribed locally on your PC and the resulting command is sent to the FiveM client console. Nothing is sent to the cloud.

Release

Which file do I download?

File	Pick this if
fivem-voicecontrols-full.streamDeckPlugin	You don't have Python. Includes its own speech engine (about 110 MB).
fivem-voicecontrols-lite.streamDeckPlugin	You already have Python 3.9+. Smaller download, but you need to install the packages yourself: pip install faster-whisper numpy sounddevice

Install

1. Download one of the files above and double-click it. Stream Deck will install the plugin.
2. Drag Voice Command from the FiveM-VoiceControls category onto a key.
3. Hold the key, speak, and release.
4. The first press downloads the speech model (about 150 MB). The key displays DOWNLOADING while this happens.

Features

• Push-to-talk speech recognition running entirely on your PC.
• Custom commands: define your own trigger phrases and FiveM console commands using JSON.
• Support for NATO phonetic alphabet, spoken numbers, and common mishearings.
• Speech corrections and custom vocabulary.
• Configurable argument types, validation, required arguments, and cooldowns.
• Configurable microphone selection.
• Transcription confidence threshold.
• Detailed parsing logs for troubleshooting.
• Key status indicators: LISTENING, THINKING, SENT, NO GAME, NOT HEARD, and error states.
• Audio is not saved unless Debug is enabled.

Known issues

• Windows only for now. The Full version is built for Windows. macOS is currently untested.
• Windows or antivirus warnings (Full version). The bundled engine is not code-signed, so Windows may show an unknown publisher warning. Select More info, then Run anyway. Some antivirus tools may also remove it. If this happens, try the Lite version.
• FiveM must be running with its console port available (default 29200), or the key will display NO GAME.

Troubleshooting

Key shows	Meaning
NO PYTHON	Lite only. Python wasn't found. Set its full path in the key's Python setting.
PIP INSTALL	Lite only. Required Python packages are missing. Run the pip install command above.
MODEL FAIL	The model download failed. Check your internet connection.
NO MIC	The microphone couldn't be opened. Check the Microphone setting and your OS microphone permissions.
NO GAME	FiveM isn't running, or the console port doesn't match.
NOT HEARD	Nothing matched a command. Enable Debug to see what was recognised.
LOW CONF	The transcription confidence was below your configured minimum.
COOLDOWN	The command was ignored because its cooldown period has not expired.
BAD CONFIG	The command configuration is invalid. Check your JSON settings.

Full documentation is available in the README.

This is an unofficial community project and is not affiliated with FiveM, Cfx.re, or Elgato.

Antivirus note

If your antivirus removes listener.exe, download listener-windows.zip from this release and extract it into the plugin's bin\listener folder, or add an exclusion for the file.

Full Changelog: v1.1.0...v1.2.1

v1.1.0

Full Changelog: v0.1.0...v1.1.0

FiveM Voice Controls v1.1.0

Hold a Stream Deck key, say a command, release. Your speech is transcribed on your PC and sent to the FiveM client console. Nothing is sent to the cloud.

Which file do I download?

File	Pick this if
fivem-voicecontrols-full.streamDeckPlugin	You don't have Python. It includes its own speech engine (about 110 MB).
fivem-voicecontrols-lite.streamDeckPlugin	You already have Python 3.9+. Tiny download, but you install the packages yourself: pip install faster-whisper numpy sounddevice

Install

1. Download one file above and double-click it. Stream Deck installs it.
2. Drag Voice Command from the FiveM-VoiceControls category onto a key.
3. Hold the key, speak, release. The first press downloads the speech model (about 150 MB). The key shows DOWNLOADING while it does.

Features

• Push-to-talk speech recognition that runs entirely on your PC
• Custom commands: define your own trigger words and console commands in the key settings (JSON)
• Handles the NATO alphabet, spoken numbers, and common mis-hearings
• Microphone picker, so you aren't tied to the system default
• Key status shows what's happening: LISTENING, THINKING, SENT, NO GAME, NOT HEARD, plus errors
• No audio is saved unless you turn on Debug

Default commands

Say	Sends
"Sam Whiskey November Tango one one"	sam WNT11
"P one two three four"	p 1234

Known issues

• Windows only for now. The Full version is built for Windows. macOS is untested.
• Windows or antivirus warnings (Full version). The bundled engine isn't code-signed. Windows may show an "unknown publisher" warning. Choose More info, then Run anyway. Some antivirus tools may remove it. If yours does, use the Lite version.
• FiveM must be running with its console port available (default 29200), or the key shows NO GAME.

Troubleshooting

Key shows	Meaning
NO PYTHON	Lite only. Python wasn't found. Set its full path in the key's Python setting.
PIP INSTALL	Lite only. Run the pip install command above.
MODEL FAIL	The model download failed. Check your internet connection.
NO MIC	The mic couldn't be opened. Check the Microphone setting and your OS microphone permissions.
NO GAME	FiveM isn't running, or the console port doesn't match.
NOT HEARD	Nothing matched a command. Turn on Debug to see what was heard.

Full docs are in the README.

This is an unofficial community project and isn't affiliated with FiveM, Cfx.re, or Elgato.

v0.1.0

Full Changelog: https://github.com/tigerjess12/fivem-voicecontrols/commits/v0.1.0

FiveM Voice Controls v0.1.0

Hold a Stream Deck key, say a command, release. Your speech is transcribed on your PC and sent to the FiveM client console. Nothing is sent to the cloud.

Which file do I download?

File	Pick this if
fivem-voicecontrols-full.streamDeckPlugin	You don't have Python. It includes its own speech engine (about 110 MB).
fivem-voicecontrols-lite.streamDeckPlugin	You already have Python 3.9+. Tiny download, but you install the packages yourself: pip install faster-whisper numpy sounddevice

Install

1. Download one file above and double-click it. Stream Deck installs it.
2. Drag Voice Command from the FiveM-VoiceControls category onto a key.
3. Hold the key, speak, release. The first press downloads the speech model (about 150 MB). The key shows DOWNLOADING while it does.

Features

• Push-to-talk speech recognition that runs entirely on your PC
• Custom commands: define your own trigger words and console commands in the key settings (JSON)
• Handles the NATO alphabet, spoken numbers, and common mis-hearings
• Microphone picker, so you aren't tied to the system default
• Key status shows what's happening: LISTENING, THINKING, SENT, NO GAME, NOT HEARD, plus errors
• No audio is saved unless you turn on Debug

Default commands

Say	Sends
"Sam Whiskey November Tango one one"	sam WNT11
"P one two three four"	p 1234

Known issues

• Windows only for now. The Full version is built for Windows. macOS is untested.
• Windows or antivirus warnings (Full version). The bundled engine isn't code-signed. Windows may show an "unknown publisher" warning. Choose More info, then Run anyway. Some antivirus tools may remove it. If yours does, use the Lite version.
• FiveM must be running with its console port available (default 29200), or the key shows NO GAME.

Troubleshooting

Key shows	Meaning
NO PYTHON	Lite only. Python wasn't found. Set its full path in the key's Python setting.
PIP INSTALL	Lite only. Run the pip install command above.
MODEL FAIL	The model download failed. Check your internet connection.
NO MIC	The mic couldn't be opened. Check the Microphone setting and your OS microphone permissions.
NO GAME	FiveM isn't running, or the console port doesn't match.
NOT HEARD	Nothing matched a command. Turn on Debug to see what was heard.

Full docs are in the README.

This is an unofficial community project and isn't affiliated with FiveM, Cfx.re, or Elgato.