diff --git a/src/voice/audio.rs b/src/voice/audio.rs
index 2e155907419..c556c8be260 100644
--- a/src/voice/audio.rs
+++ b/src/voice/audio.rs
@@ -6,7 +6,6 @@ use std::{
 
 pub const HEADER_LEN: usize = 12;
 pub const SAMPLE_RATE: u32 = 48_000;
-pub static SILENT_FRAME: [u8; 3] = [0xf8, 0xff, 0xfe];
 
 /// A readable audio source.
 pub trait AudioSource: Send {
diff --git a/src/voice/connection.rs b/src/voice/connection.rs
index 54d91163dfa..d1472d5e667 100644
--- a/src/voice/connection.rs
+++ b/src/voice/connection.rs
@@ -45,7 +45,7 @@ use std::{
     },
     time::Duration
 };
-use super::audio::{AudioReceiver, AudioType, LockedAudio, HEADER_LEN, SAMPLE_RATE, SILENT_FRAME};
+use super::audio::{AudioReceiver, AudioType, HEADER_LEN, SAMPLE_RATE, LockedAudio};
 use super::connection_info::ConnectionInfo;
 use super::{payload, VoiceError, CRYPTO_MODE};
 use websocket::{
@@ -200,8 +200,7 @@ impl Connection {
             keepalive_timer: Timer::new(temp_heartbeat),
             udp,
             sequence: 0,
-            // We need to send some frames to receive any audio.
-            silence_frames: 100,
+            silence_frames: 0,
             soft_clip,
             speaking: false,
             ssrc: hello.ssrc,
@@ -393,7 +392,7 @@ impl Connection {
                 self.silence_frames -= 1;
 
                 // Explicit "Silence" frame.
-                opus_frame.extend_from_slice(&SILENT_FRAME);
+                opus_frame.extend_from_slice(&[0xf8, 0xff, 0xfe]);
             } else {
                 // Per official guidelines, send 5x silence BEFORE we stop speaking.
                 self.set_speaking(false)?;