Skip to content
Version 2.32.0
↓ Get the launcher

IVoiceInjection interface

Injects audio into the game’s OUTGOING team-voice stream (mixed with the live microphone). Backed by a process-wide native DirectSound capture hook installed by the framework at boot. Multiple sources — across all plugins — mix together into the one shared uplink.

public interface IVoiceInjection
name description
Available { get; } True once the capture hook is installed AND the capture format is known (i.e. the mic is live, post-login). Audio assigned to a source before this is retained and becomes audible once true; it is simply not mixed yet. Poll or gate playback on this if you need the audio heard immediately. Read any thread; it reflects the most recent capture-format observation.
Status { get; } Hook / capture-format diagnostics for status UIs and logs. Read any thread (a snapshot).
CreateSource() Create an independent playback voice in the mixer. Returns null if the mixer is at capacity. Dispose it to remove it from the mixer (sources a plugin leaves open are auto-disposed on plugin teardown). Typically a plugin keeps one source for looped background audio and/or uses PlayOnce for one-shots. Main-thread only.
LoadFileAsync(…) Decode an audio file off-thread (WAV natively; MP3/WMA/M4A via Windows Media Foundation) into a reusable clip. Does NOT require the capture format to be known — the clip keeps its source PCM and is (re)resampled to the capture format on demand. Throws on a missing file or an unsupported/corrupt encoding. Media Foundation is absent on Windows “N” editions → non-WAV will fail there (WAV still works). Call from any thread; the returned task completes on a background thread.
LoadPcm(…) Build a clip from caller-supplied 16-bit interleaved PCM at the given sample rate and channel count. Copied immediately (the span need not outlive the call); resampled to the capture format on demand. Use for TTS or generated audio. Main-thread safe; returns at once (no decode).
SetUplinkPolicy(…) Configure the GVoice uplink DSP, auto-applied while ANY source is playing and restored to the game’s defaults when all sources stop. GVoice’s noise-suppression / AGC attenuate injected (non-voice) audio, so disabling them preserves music fidelity; a higher Opus bitrate improves quality (up to the 64 kbps ceiling). The uplink encoder is process-global, so this is a single global policy — the last writer wins across plugins. A null argument leaves that setting unchanged. A disposed plugin’s policy becomes inert once its sources stop. Main-thread only (never called from the native mix hook).

The uplink encoder and the capture hook are process-global, so some state here is framework-wide (the DSP policy, the capture format) while each IVoiceSource is an independent voice. Sources a plugin creates are auto-disposed when the plugin unloads (mirroring IHarmonyHost auto-unpatch), so a plugin cannot leak a voice across a soft enable/disable cycle.