Skip to main content

Changelog

2.0.0 – 04 August 2026 (UE 5.5+)​

This is a major update focused on voice quality, microphone reliability, packet loss recovery, and platform support. 2.0.0 is available on Fab.com as of UE 5.5 but the plugin also supports earlier engine versions. Try to use it as a project plugin and contact me if you need any help.

Audio Capture​

Microphone capture now uses the engine's AudioCapture module instead of the limited IVoiceCapture interface. AudioCapture is the engine's modern interface for capturing microphones on any platform. This opens a lot more possibilities on platform support and OS microphone control. iOS is now also officially supported. Consoles should now already work or be possible to support with little effort. But I am unable to provide further details because I do not have access to the console SDKs.

You can now get the input device currently used for voice capture with the new function UOnsetVoipLocalPlayerSubsystem::GetVoiceCaptureDevice. The new delegate OnVoipCaptureLost is called if voice capture is ever lost during capturing. For example when unplugging the microphone. The new delegate OnVoipCaptureStateChange is called whenever voice capture starts or stops, matching IsCapturingVoice(). Changing Windows microphone settings at runtime is supported. The device will properly recover and not stay in a broken state.

GetCurrentAmplitude() now works on any platform.

Noise Suppression​

Using the RNNoise library. This library uses a neural network to extract background noise from voice. If you wish, noise suppression can be disabled in the VoIP settings in the editor. It can also be compiled out by setting bEnableRNNoise = false in OnsetVoip.Build.cs.

Voice Encoding​

The latest libopus library (version 1.6.1) is now bundled with the plugin. Before, it was using the engine's libopus library which is several years old. This library is used for encoding voice data to be sent over the network. Using the latest version alone brings dramatic performance improvements. The plugin now also makes use of the features this brilliant library provides:

  • Forward Error Correction (FEC): If voice data is ever lost over the network, FEC will help recover it resulting in smoother voice quality without gaps.
  • Packet Loss Concealment (PLC): Similar to FEC but it uses a neural network (Deep PLC) to recover lost audio data. The plugin will either choose FEC or PLC.
  • Discontinuous Transmission (DTX): An additional layer to avoid sending "silence" over the network. The algorithm detects whether there is voice or not. This works together with RNNoise.

Prebuilt libraries are provided but if you still wish to use the engine's library you can set bUseBundledOpus = false in OnsetVoip.Build.cs.

Improved network resilience​

The new jitter buffer and other improvements in the code now better handle different network conditions like lag and packet loss.

Speaker components​

Added UOnsetVoipSpeakerComponent which enables you to easily build a speaker system on the map. The updated example project on GitHub contains a demonstration of it.

speaker component demo

Place or spawn a UOnsetVoipSpeakerComponent on the map and call SetTalkerForPlayback, the plugin will take care of the rest.

Automatic Gain Control​

The plugin now properly levels your microphone input volume resulting in a smooth playback for remote players even if you talk quietly or loudly. (voice.MicAutoLevel)

Pitch Shift Effect​

Added UOnsetSourceEffectPitchShiftPreset, a source effect for changing voice pitch without speeding up or slowing down procedural audio playback.

Reduced setup complexity​

Zero Config Setup. No more project config is required. Just enable the plugin, restart the editor and you are good to go. Setting ChannelDefinitions is still supported as it was before. This is for users who have very special net drivers.

Setting the microphone sample rate and channels is no longer needed. AudioCapture picks up the device settings automatically.

Setting the microphone sensitivity using CVars is no longer needed. The plugin automatically handles that.

I tested the plugin with UE 5.5 and 5.7 on:

  • Windows 11
  • Ubuntu 24.04
  • iOS
  • Android
  • macOS
  • Pixel Streaming platforms

The settings under Edit -> Project Settings -> Onset VoIP Settings now allow setting default USoundClass values separately for 2D and 3D voice audio components with DefaultSoundClass2D and DefaultSoundClass3D. You no longer have to override the classes for this.

Upgrading from 1.8.5 to 2.0.0:​

This should cover all changes that you might want to be aware of when upgrading from the earlier version.

If you were using the old default bitrate of 24000, I recommend using 36000 for 2.0.0. FEC is enabled by default and benefits from the higher bitrate.

  • Removed CancelReceivedAudioData(). Use UOnsetVoipTalker::SetVoipEnabled or set component volume to 0.
  • Audio pipeline is now fully float instead of int16. Therefore delegates like OnVoipMicrophoneAudioCaptured now deliver float instead of int values.
  • Voice packet/network format changed, so client and server should be updated to 2.0.0 together.
  • Use the new CVar voice.MicGain instead of the old voice.MicInputGain/voice.MicInputGain2.
  • Use the new CVar voice.MicAutoLevel instead of the old voice.MicInputAGC.
  • The old CVars to control microphone "sensitivity" (which were part of VoiceCapture module) are no longer needed: voice.MicNoiseGateThreshold and voice.SilenceDetectionThreshold. The plugin automatically detects when the microphone contains speech. If you still wish to influence that process there is voice.VoiceGateBias, but this should not be required in most cases.
  • The Blueprint mic helper functions SetMicNoiseGateThreshold, GetMicNoiseGateThreshold, SetSilenceDetectionThreshold, and GetSilenceDetectionThreshold were removed.
  • The audio input device helper changed: UAudioCaptureBlueprintLibrary2 and FAudioInputDeviceInfo2 were removed. Use UOnsetAudioCaptureBlueprintLibrary and Unreal's FAudioInputDeviceInfo instead.
  • Old config/settings no longer apply: VoiceCaptureSampleRate, VoiceCaptureChannels, DefaultSoundWaveVolume, EncodingHint, bAdvancedEncoderSettings, and bEnableVBR.
  • DefaultSoundWaveVolume was removed. Remote voice volume is now controlled on the audio component, for example through voice.volume2d, voice.volume3d, or the component volume.
  • UOnsetVoipSoundWave::GetAttenuation() and UOnsetVoipSoundWave::SetAttenuation() were removed. Set attenuation and source effects on UOnsetVoipAudioComponent instead.
  • If you customize voice relevancy in C++, VoipNetRelevancy is now on UOnsetVoipWorldSubsystem instead of UOnsetVoipDataChannel.
  • Recordings now use the new fixed 24 kHz mono voice pipeline. Raw PCM recordings are still signed 16-bit PCM, but no longer use VoiceCaptureSampleRate or VoiceCaptureChannels.
  • Recordings without a known file extension now default to WAV instead of raw PCM.
  • Distance-based relevancy is now enabled by default.
  • Settings changes:
    • NumberOfFramesToStopPlaybackIfNoAudio -> TimeToStopPlaybackIfNoAudio
    • Network jitter buffering is now controlled by InitialJitterBufferDelay and MaxJitterBufferDelay
    • Decoded PCM playback buffering is now controlled separately by InitialPlaybackBufferDelay and MaxPlaybackBufferDelay

1.8.5 – 25 May 2026 (UE 5.5+)​

Version 2 of this plugin is under development and can be beta tested. If you are interested join the new Discord server: https://discord.gg/myUmdZddgx

  • Improve voice audio playback in different network situations.
  • Add InitialPlaybackBufferDelay to the settings menu. Buffered amount of voice before playback.
  • Fix support for spectators.
  • Android: Fix a crash on capture shutdown.
  • Android: Fix voice capture if permission is not granted on plugin startup. It will retry on capture start.
  • Other minor fixes.

Don’t forget to check out the new Voice Example project.

BlueMountainsIO/OnsetVoipExample

1.8.4 – 25 January 2026 (UE 5.5+)​

  • Fix amplitude replication
  • Fix listen server not able to hear other players sometimes
  • Renamed function in UOnsetVoipTalker ProcessVoiceOverrideRelevancy to ProcessVoiceOverrideRelevancy
  • Do not send voice to new connections that are not aware of the player state yet.
  • Minor code improvements

1.8.3 – 04 December 2025 (UE 5.5+, backported for UE 5.4)​

  • Fix a crash when joining a session with active speakers.

1.8.2 – 18 November 2025 (UE 5.5+)​

  • Android: Fix static noise on capture start. This is a workaround because of an engine bug.
  • Fix two compile warnings for UE 5.6+
  • Empty voice remainder before capture start.
  • Minor code improvements.

1.8.1 – 02 November 2025 (UE 5.4+)​

  • When calling ToggleVoiceCapture(false) and then immediately afterwards ToggleVoiceCapture(true) the device might not have fully stopped resulting in the device not being started. This case is now intercepted and capture is started once device stopped. Unless ToggleVoiceCapture(false) is called in between. This is non-blocking and happens mostly within 20 milliseconds.
  • Fix a rare crash when voice tries to play on an audio component without a sound wave being initialized.

1.8 – 10 October 2025 (UE 5.4+)​

  • Added the feature to play 2D voice channels in proximity (3D) if the other player is nearby. This is done by enabling UOnsetVoipTalker::bEnablePrefer2DVoicePlaybackOn3DIfNear Useful for cases where you use voice channels for walkie talkies but want to rather hear the other player in 3D if they are near. If the other player is not near then it will play in 2D.
  • Added the feature to play 2D voice channels in proximity (3D) and 2D at the same time. Some people have requested this feature. This can be enabled with UOnsetVoipTalker::bEnable2DVoicePlaybackOnBoth2DAnd3D Voice will still only be sent once to the client to save bandwidth.
  • Boolean variables like bIs2D to differentiate between proximity and 2D voice channels have been replaced by EOnsetVoipNetRelevancy. If you get errors when compiling blueprints, please update to the new signatures. This also includes delegates like OnVoipTalkingStateChange.
  • UOnsetVoipTalker::IsLastReceivedTalkingState2D was replaced by GetLastVoipRelevancy
  • Added virtual function UOnsetVoipTalker::ProcessVoiceOverrideRelevancy. This is a good place to override relevancy on the client just before playback.
  • Added UOnsetVoipSettings::DefaultOnsetVoipAudioComponentClass2D/3D. Until now you would need to also subclass UOnsetVoipTalker to change the default audio component class. Now you can set it under Edit -> Project Settings -> Onset VoIP Settings. The same variables remain in the UOnsetVoipTalker class as well and have a higher priority.
  • Until now UOnsetVoipTalker::GetVoiceAudioComponent would return null if the talker did not receive any voice yet. Now an audio component will be created on call if it doesn’t exist and saved/cached internally for playback.

1.7.1 – 15 September 2025​

  • Fix a crash during seamless travel that may happen sometimes.

1.7 – 01 May 2025​

  • Add initial support for PixelStreaming.
  • Fix disconnects on seamless travel.
  • Fix nameplate still showing when pawn possession changes.
  • The example project has been recreated from scratch. It now includes more usage of existing plugin features.
  • SetNameplateSpeaking and SetNameplatePlayerName has been removed from IOnsetVoipNameWidgetInterface. SetOwningPlayerState will be called with the relevant player state that you can use to get names and talking state. See the new example project on how it is done in WBP_Nameplate.
  • IOnsetVoipPawnInterface::OnNewPlayerState has been renamed to OnVoipTalkerChange. After the pawn changes for a player state this interface function is called on the old pawn with a null value as parameter. This interface function is now also called on dedicated servers.
  • Added UOnsetVoipAudioComponent2D that is a child class of UOnsetVoipAudioComponent. This is now used as a base for 2D audio (from voice channels). Affects setting UOnsetVoipTalker::DefaultOnsetVoipAudioComponentClass2D
  • Add two new CVars that control the default audio volume multiplier for voice audio components (other players speaking). These are voice.volume2d for voice channel 2D audio and voice.volume3d for proximity (channel 0)
  • Add static functions to UOnsetVoipBlueprintLibrary SetVoipAudioComponentsVolume_2DVoice, SetVoipAudioComponentsVolume_3DVoice SetVoipAudioComponentsVolume GetPIEInstanceID PrintToConsole GetOnsetVoipVersion SetMicNoiseGateThreshold SetSilenceDetectionThreshold
  • Minor code improvements

1.6.1 – 23 February 2025​

  • Fix an issue with packet serialization resulting on disconnects on certain platforms.

1.6 – 18 November 2024​

  • Fully rewrote the voice recording feature. It now supports recording remote players to a single audio file, .wav files, conversion to mp3 using ffmpeg and recording/decoding voice on dedicated servers.
  • To enable voice processing on a dedicated server set bEnableVoiceProcessing to true in the OnsetVoip.Build.cs file. This will enable an entire set of features on dedicated servers, all listed inside the OnsetVoip.Build.cs file.
  • Add functions UOnsetVoipWorldSubsystem::StartRecording, StopRecording, IsRecording and GetRecordingFile().
  • Add functions UOnsetVoipLocalPlayerSubsystem::StartRecording, StopRecording, IsRecording and GetRecordingFile().
  • Add console command voice.device.
  • Microphone amplitude is now synced. It can be retrieved for remote players with UOnsetVoipTalker::GetCurrentAmplitude()
  • Moved OnsetVoipPacket.h and OnsetVoipDataChannel.h to public source folder.
  • Made functions in FOnsetVoipNetRelevancy virtual and able to override in UOnsetVoipDataChannel::PostInitProperties()
  • Add pure C++ delegate, Native_OnVoipTalkerCreated and Native_OnVoipTalkerDestroyed mimicking the BP delegates.
  • Fix a crash related to nameplates on certain ACharacter classes that return null in GetMesh().
  • Fix a bug in IsVoiceWorldRelevant that caused wrong replication of some players.
  • Improved some comments.

1.5 – 22 September 2024​

  • Add console command voice.record.local, voice.record.remote to record local microphone or voice from remote UOnsetVoipTalker. Output location is printed in the logs, file format is signed 16-bit PCM with channels and sample rate as defined under Edit -> Project Settings -> Onset VoIP Settings. The file can be opened with programs like Audacity.
  • Add console command voice.loopback that triggers UOnsetLocalPlayerSubsystem::ToggleLoopback.
  • Add temporary/experimental setting NumberOfFramesToStopPlaybackIfNoAudio that defines the number of frames until a voice audio component stops playing due to no playback data.
  • Only queue audio to procedural sound wave if there are at least InSamplesToGenerate in OnGenerateProceduralAudio().
  • Expose opus encoder setting bEnableVBR (Variable Bitrate) in the plugins project setting.
  • Add log category LogOnsetVoipPacket for serialization errors of FOnsetVoipPacket.
  • Print debug log messages about some voice CVars.
  • Add plugin project setting DefaultSoundWaveVolume that sets the default volume of remote talkers.
  • Packet voice data checksum is now enabled by default. To disable set the compile time constant bEnableVoiceDatChecksum = false.
  • Minor code improvements and fixes

1.4 – 01 August 2024​

  • Fix nameplate widget flickering in low FPS situations.
  • Add a limit on how many seconds of voice data may be buffered before playback must occur. Value can be changed under Edit -> Project Settings -> Onset VoIP Settings -> Advanced -> Max Buffer Playback Delay
  • Add UOnsetVoipLocalPlayerSubsystem::ToggleLoopback to be able to hear yourself.
  • Experimental CVar voice.MicInputAGC Automatic Gain Control feature to normalize microphone input gain.
  • Experimental distance-based network relevancy checks. Edit -> Project Settings -> Onset VoIP Settings -> Advanced -> Use Distance Based Relevancy
  • Minor fixes.

1.3 – 04 July 2024​

  • Add support for the client telling the server on what voice channels we want to speak in. Useful for features where you have a keyboard shortcut for talking in the world channel and another shortcut for talking inside a voice channel. The new functions are ToggleVoiceCaptureWithChannel and ToggleVoiceCaptureWithChannels. The talker must be part of the requesting voice channel on the server. See UOnsetVoipTalker::SetVoiceChannel.
  • Smaller fixes and improvements.

1.2 – 25 June 2024​

  • Add console variable voice.MicInputGain2 to change the volume of the local microphone recording. The original engine function voice.MicInputGain should not be used as it causes voice artifacts.
  • Add function for retrieving microphone input loudness between 0.0-1.0: UOnsetVoipLocalPlayerSubsystem::GetCurrentAmplitude()
  • Change plugin loading phase to PreDefault.
  • Move stat voip code to UOnsetStatVoipSubsystem because GEngine is not valid in StartupModule with PreDefault loading phase.
  • Add C++ delegate for capturing local microphone input: UOnsetVoipLocalPlayerSubsystem::Native_OnVoipMicrophoneAudioCaptured
  • Add function for retrieving all UOnsetVoipTalker in a specific voice channel: UOnsetVoipWorldSubsystem::GetAllTalkersInVoiceChannel()
  • Fix decompression buffer for very high voice capture quality settings.
  • You can now derive your own child class from UOnsetVoipTalker and set it in Edit -> Project Settings -> Onset VoIP Settings -> Default Onset VoIP Talker Class (DefaultOnsetVoipTalkerClass).
  • Made functions virtual in UOnsetVoipTalker so they can be overridden.
  • You can now derive your own child class from UOnsetVoipAudioComponent and set it in a custom UOnsetVoipTalker class for both 2D and 3D voice.
  • Add UOnsetVoipAudioComponent::DefaultOnsetVoipSoundWaveClass to change it to a custom child class derived from UOnsetVoipSoundWave.
  • Add delegates to intercept audio data for UOnsetVoipTalker when received. Format is raw signed 16-bit PCM. Number of channels and sample rate are defined in UOnsetVoipSettings::VoiceCaptureChannels and VoiceCaptureSampleRate. UOnsetVoipTalker::OnVoipAudioDataReceived UOnsetVoipTalker::Native_OnVoipAudioDataReceived UOnsetVoipWorldSubsystem::OnVoipAudioDataReceived UOnsetVoipWorldSubsystem::Native_OnVoipAudioDataReceived
  • Add UOnsetVoipTalker::CancelReceivedAudioData() that can be called during one of the above callbacks to stop the plugin from further processing the audio data (i.e. playing the audio).
  • Add functions to mute a talker on the server side for another talker: UOnsetVoipTalker::SetMutedForPlayerState, SetMutedForVoipTalker, IsMutedForVoipTalker and IsMutedForPlayerState
  • Add functions to mute a talker on a voice channel. UOnsetVoipTalker::SetMutedOnVoiceChannel, IsMutedOnVoiceChannel
  • Add a way to do custom logic for sending voice data to a specific talker: UOnsetVoipTalker::ShouldSendVoiceTo(UOnsetVoipTalker* OtherVoipTalker)
  • Add UOnsetVoipBlueprintLibrary for exposing some functions to blueprints without getting a subsystem first. Was requested by a user and does not contain all functions yet.
  • Add "Client/Server/Shared" to function category names so you more easily see the scope of the function.

1.1 – 15 June 2024​

  • Add BitRate setting for the libopus encoder under Edit -> Project Settings -> Onset VoIP Settings -> Bit Rate.
  • Fixed some content assets and default nameplate setting.
  • Fixed a crash on pawn possession.

1.0 – 13 June 2024​

  • Initial release.