Use the same opus lib for encoding as decoding. Simplify by encoding float64 directly.

speaker/discord.go

@@ -1,8 +1,6 @@
 package discordspeaker
 
 import (
-	"bytes"
-	"encoding/binary"
 	"log"
 	"sync"
 	"time"
@@ -10,7 +8,7 @@ import (
 	"github.com/bwmarrin/discordgo"
 	"github.com/faiface/beep"
 	"github.com/pkg/errors"
-	"layeh.com/gopus"
+	"gopkg.in/hraban/opus.v2"
 )
 
 var (
@@ -18,15 +16,13 @@ var (
 	mixer      beep.Mixer
 	samples    [][2]float64
 	done       chan struct{}
-	encoder    *gopus.Encoder
+	encoder    *opus.Encoder
 	voice      *discordgo.VoiceConnection
 	frameSize  int = 960
 	channels   int = 2
 	sampleRate int = 48000
 	maxBytes   int = (frameSize * 2) * 2
-	pcm        []int16
 	buf        []byte
-	pause      bool
 )
 
 func Init(dgv *discordgo.VoiceConnection) error {
@@ -35,17 +31,15 @@ func Init(dgv *discordgo.VoiceConnection) error {
 	mu.Lock()
 	defer mu.Unlock()
 	Close()
+
 	mixer = beep.Mixer{}
-
 	buf = make([]byte, maxBytes)
-	pcm = make([]int16, frameSize*channels)
 	samples = make([][2]float64, frameSize)
 
-	pause = true
-
 	voice = dgv
 
-	encoder, err = gopus.NewEncoder(sampleRate, channels, gopus.Audio)
+	encoder, err = opus.NewEncoder(sampleRate, channels, opus.AppVoIP)
+	encoder.SetBitrateToMax()
 
 	if err != nil {
 		return errors.Wrap(err, "failed to initialize speaker")
@@ -93,34 +87,24 @@ func update() {
 	mu.Lock()
 	mixer.Stream(samples)
 	mu.Unlock()
+	var f32 []float32
 
-	for i := range samples {
+	for _, sample := range samples {
-		for c := range samples[i] {
+		for _, val := range sample {
-			val := samples[i][c]
+			f32 = append(f32, float32(val))
-			if val < -1 {
-				val = -1
-			}
-			if val > +1 {
-				val = +1
-			}
-			valInt16 := int16(val * (1<<15 - 1))
-			low := byte(valInt16)
-			high := byte(valInt16 >> 8)
-			buf[i*4+c*2+0] = low
-			buf[i*4+c*2+1] = high
-		}
 		}
 
-	if Silence(buf) {
+	}
+
+	if Silence(f32) {
 		time.Sleep(100 * time.Millisecond)
 		return
 	}
 
-	binary.Read(bytes.NewReader(buf), binary.LittleEndian, &pcm)
+	n, err := encoder.EncodeFloat32(f32, buf)
-
-	opus, err := encoder.Encode(pcm, frameSize, maxBytes)
 	if err != nil {
 		log.Println(err)
+		time.Sleep(100 * time.Millisecond)
 		return
 	}
 
@@ -128,10 +112,10 @@ func update() {
 		return
 	}
 
-	voice.OpusSend <- opus
+	voice.OpusSend <- buf[:n]
 }
 
-func Silence(in []byte) bool {
+func Silence(in []float32) bool {
 	for _, v := range in {
 		if v != 0 {
 			return false