feat: Implement TeamSpeak 3 client with connection management, event handling, and audio support.
This commit is contained in:
Jose Luis Montañes Ojados
261
pkg/audio/capture.go
Normal file
261
pkg/audio/capture.go
Normal file
@@ -0,0 +1,261 @@
|
||||
package audio
|
||||
|
||||
import (
|
||||
"encoding/binary"
|
||||
"fmt"
|
||||
"sync"
|
||||
"time"
|
||||
"unsafe"
|
||||
|
||||
"github.com/go-ole/go-ole"
|
||||
"github.com/moutend/go-wca/pkg/wca"
|
||||
)
|
||||
|
||||
const captureFrameSamples = 960 // 20ms at 48kHz
|
||||
|
||||
// Capturer handles WASAPI audio capture from microphone
|
||||
type Capturer struct {
|
||||
client *wca.IAudioClient
|
||||
captureClient *wca.IAudioCaptureClient
|
||||
waveFormat *wca.WAVEFORMATEX
|
||||
bufferSize uint32
|
||||
running bool
|
||||
mu sync.Mutex
|
||||
stopChan chan struct{}
|
||||
|
||||
// Callback for captured audio (called with 960-sample frames)
|
||||
onAudio func(samples []int16)
|
||||
|
||||
// Sample accumulation buffer
|
||||
sampleBuffer []int16
|
||||
bufferMu sync.Mutex
|
||||
|
||||
// Current audio level (0-100)
|
||||
currentLevel int
|
||||
levelMu sync.RWMutex
|
||||
}
|
||||
|
||||
// NewCapturer creates a new WASAPI audio capturer
|
||||
func NewCapturer() (*Capturer, error) {
|
||||
// Initialize COM (may already be initialized)
|
||||
ole.CoInitializeEx(0, ole.COINIT_APARTMENTTHREADED)
|
||||
|
||||
// Get default capture endpoint (microphone)
|
||||
var deviceEnumerator *wca.IMMDeviceEnumerator
|
||||
if err := wca.CoCreateInstance(
|
||||
wca.CLSID_MMDeviceEnumerator,
|
||||
0,
|
||||
wca.CLSCTX_ALL,
|
||||
wca.IID_IMMDeviceEnumerator,
|
||||
&deviceEnumerator,
|
||||
); err != nil {
|
||||
return nil, fmt.Errorf("failed to create device enumerator: %w", err)
|
||||
}
|
||||
defer deviceEnumerator.Release()
|
||||
|
||||
var device *wca.IMMDevice
|
||||
if err := deviceEnumerator.GetDefaultAudioEndpoint(wca.ECapture, wca.EConsole, &device); err != nil {
|
||||
return nil, fmt.Errorf("failed to get default capture device: %w", err)
|
||||
}
|
||||
defer device.Release()
|
||||
|
||||
// Activate audio client
|
||||
var audioClient *wca.IAudioClient
|
||||
if err := device.Activate(wca.IID_IAudioClient, wca.CLSCTX_ALL, nil, &audioClient); err != nil {
|
||||
return nil, fmt.Errorf("failed to activate audio client: %w", err)
|
||||
}
|
||||
|
||||
// Set up format for 48kHz mono 16-bit (TeamSpeak format)
|
||||
waveFormat := &wca.WAVEFORMATEX{
|
||||
WFormatTag: wca.WAVE_FORMAT_PCM,
|
||||
NChannels: 1,
|
||||
NSamplesPerSec: 48000,
|
||||
WBitsPerSample: 16,
|
||||
NBlockAlign: 2,
|
||||
NAvgBytesPerSec: 96000,
|
||||
CbSize: 0,
|
||||
}
|
||||
|
||||
// Initialize in shared mode - 100ms buffer
|
||||
duration := wca.REFERENCE_TIME(100 * 10000) // 100ms in 100-nanosecond units
|
||||
if err := audioClient.Initialize(
|
||||
wca.AUDCLNT_SHAREMODE_SHARED,
|
||||
wca.AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM|wca.AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY,
|
||||
duration,
|
||||
0,
|
||||
waveFormat,
|
||||
nil,
|
||||
); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to initialize audio client: %w", err)
|
||||
}
|
||||
|
||||
// Get buffer size
|
||||
var bufferSize uint32
|
||||
if err := audioClient.GetBufferSize(&bufferSize); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to get buffer size: %w", err)
|
||||
}
|
||||
|
||||
// Get capture client
|
||||
var captureClient *wca.IAudioCaptureClient
|
||||
if err := audioClient.GetService(wca.IID_IAudioCaptureClient, &captureClient); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to get capture client: %w", err)
|
||||
}
|
||||
|
||||
return &Capturer{
|
||||
client: audioClient,
|
||||
captureClient: captureClient,
|
||||
waveFormat: waveFormat,
|
||||
bufferSize: bufferSize,
|
||||
stopChan: make(chan struct{}),
|
||||
sampleBuffer: make([]int16, 0, captureFrameSamples*50), // ~1 second buffer
|
||||
}, nil
|
||||
}
|
||||
|
||||
// SetCallback sets the callback for captured audio (receives 960-sample frames)
|
||||
func (c *Capturer) SetCallback(fn func(samples []int16)) {
|
||||
c.mu.Lock()
|
||||
c.onAudio = fn
|
||||
c.mu.Unlock()
|
||||
}
|
||||
|
||||
// Start begins audio capture
|
||||
func (c *Capturer) Start() error {
|
||||
c.mu.Lock()
|
||||
if c.running {
|
||||
c.mu.Unlock()
|
||||
return nil
|
||||
}
|
||||
c.running = true
|
||||
c.stopChan = make(chan struct{}) // Recreate channel for each start
|
||||
c.mu.Unlock()
|
||||
|
||||
if err := c.client.Start(); err != nil {
|
||||
return fmt.Errorf("failed to start audio client: %w", err)
|
||||
}
|
||||
|
||||
go c.captureLoop()
|
||||
return nil
|
||||
}
|
||||
|
||||
// Stop stops audio capture
|
||||
func (c *Capturer) Stop() {
|
||||
c.mu.Lock()
|
||||
if !c.running {
|
||||
c.mu.Unlock()
|
||||
return
|
||||
}
|
||||
c.running = false
|
||||
c.mu.Unlock()
|
||||
|
||||
close(c.stopChan)
|
||||
c.client.Stop()
|
||||
}
|
||||
|
||||
// Close releases all resources
|
||||
func (c *Capturer) Close() {
|
||||
c.Stop()
|
||||
if c.captureClient != nil {
|
||||
c.captureClient.Release()
|
||||
}
|
||||
if c.client != nil {
|
||||
c.client.Release()
|
||||
}
|
||||
}
|
||||
|
||||
// GetLevel returns the current audio input level (0-100)
|
||||
func (c *Capturer) GetLevel() int {
|
||||
c.levelMu.RLock()
|
||||
defer c.levelMu.RUnlock()
|
||||
return c.currentLevel
|
||||
}
|
||||
|
||||
// IsRunning returns whether capture is active
|
||||
func (c *Capturer) IsRunning() bool {
|
||||
c.mu.Lock()
|
||||
defer c.mu.Unlock()
|
||||
return c.running
|
||||
}
|
||||
|
||||
func (c *Capturer) captureLoop() {
|
||||
ticker := time.NewTicker(10 * time.Millisecond) // Check more often than 20ms
|
||||
defer ticker.Stop()
|
||||
|
||||
for {
|
||||
select {
|
||||
case <-c.stopChan:
|
||||
return
|
||||
case <-ticker.C:
|
||||
c.readFromBuffer()
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (c *Capturer) readFromBuffer() {
|
||||
// Read all available packets
|
||||
for {
|
||||
var packetLength uint32
|
||||
if err := c.captureClient.GetNextPacketSize(&packetLength); err != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if packetLength == 0 {
|
||||
break
|
||||
}
|
||||
|
||||
var buffer *byte
|
||||
var numFrames uint32
|
||||
var flags uint32
|
||||
if err := c.captureClient.GetBuffer(&buffer, &numFrames, &flags, nil, nil); err != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if numFrames == 0 {
|
||||
c.captureClient.ReleaseBuffer(numFrames)
|
||||
continue
|
||||
}
|
||||
|
||||
samples := make([]int16, numFrames)
|
||||
bufSlice := unsafe.Slice(buffer, numFrames*2)
|
||||
|
||||
for i := uint32(0); i < numFrames; i++ {
|
||||
samples[i] = int16(binary.LittleEndian.Uint16(bufSlice[i*2:]))
|
||||
}
|
||||
|
||||
c.captureClient.ReleaseBuffer(numFrames)
|
||||
|
||||
// Skip silent buffers
|
||||
if flags&wca.AUDCLNT_BUFFERFLAGS_SILENT != 0 {
|
||||
continue
|
||||
}
|
||||
|
||||
// Add to sample buffer
|
||||
c.bufferMu.Lock()
|
||||
c.sampleBuffer = append(c.sampleBuffer, samples...)
|
||||
|
||||
// Calculate level from latest samples
|
||||
level := CalculateRMSLevel(samples)
|
||||
c.levelMu.Lock()
|
||||
c.currentLevel = level
|
||||
c.levelMu.Unlock()
|
||||
|
||||
// Send complete 960-sample frames
|
||||
for len(c.sampleBuffer) >= captureFrameSamples {
|
||||
frame := make([]int16, captureFrameSamples)
|
||||
copy(frame, c.sampleBuffer[:captureFrameSamples])
|
||||
c.sampleBuffer = c.sampleBuffer[captureFrameSamples:]
|
||||
|
||||
// Call callback with the frame
|
||||
c.mu.Lock()
|
||||
callback := c.onAudio
|
||||
c.mu.Unlock()
|
||||
|
||||
if callback != nil {
|
||||
callback(frame)
|
||||
}
|
||||
}
|
||||
c.bufferMu.Unlock()
|
||||
}
|
||||
}
|
||||
100
pkg/audio/level.go
Normal file
100
pkg/audio/level.go
Normal file
@@ -0,0 +1,100 @@
|
||||
package audio
|
||||
|
||||
import (
|
||||
"math"
|
||||
)
|
||||
|
||||
// CalculateRMSLevel calculates the RMS level of PCM samples and returns 0-100
|
||||
func CalculateRMSLevel(samples []int16) int {
|
||||
if len(samples) == 0 {
|
||||
return 0
|
||||
}
|
||||
|
||||
var sum float64
|
||||
for _, s := range samples {
|
||||
sum += float64(s) * float64(s)
|
||||
}
|
||||
|
||||
rms := math.Sqrt(sum / float64(len(samples)))
|
||||
// Normalize to 0-100 (max int16 is 32767)
|
||||
level := int(rms / 32767.0 * 100.0)
|
||||
if level > 100 {
|
||||
level = 100
|
||||
}
|
||||
return level
|
||||
}
|
||||
|
||||
// CalculatePeakLevel returns the peak level of PCM samples as 0-100
|
||||
func CalculatePeakLevel(samples []int16) int {
|
||||
if len(samples) == 0 {
|
||||
return 0
|
||||
}
|
||||
|
||||
var peak int16
|
||||
for _, s := range samples {
|
||||
if s < 0 {
|
||||
s = -s
|
||||
}
|
||||
if s > peak {
|
||||
peak = s
|
||||
}
|
||||
}
|
||||
|
||||
return int(float64(peak) / 32767.0 * 100.0)
|
||||
}
|
||||
|
||||
// LevelToBar converts a 0-100 level to a visual bar string
|
||||
func LevelToBar(level, width int) string {
|
||||
if level < 0 {
|
||||
level = 0
|
||||
}
|
||||
if level > 100 {
|
||||
level = 100
|
||||
}
|
||||
|
||||
filled := level * width / 100
|
||||
empty := width - filled
|
||||
|
||||
bar := ""
|
||||
for i := 0; i < filled; i++ {
|
||||
bar += "█"
|
||||
}
|
||||
for i := 0; i < empty; i++ {
|
||||
bar += "░"
|
||||
}
|
||||
|
||||
return bar
|
||||
}
|
||||
|
||||
// LevelToMeter converts a 0-100 level to a visual VU meter with varying heights
|
||||
func LevelToMeter(level, width int) string {
|
||||
if level < 0 {
|
||||
level = 0
|
||||
}
|
||||
if level > 100 {
|
||||
level = 100
|
||||
}
|
||||
|
||||
// Use block characters of varying heights
|
||||
blocks := []rune{'░', '▁', '▂', '▃', '▄', '▅', '▆', '▇', '█'}
|
||||
|
||||
meter := ""
|
||||
for i := 0; i < width; i++ {
|
||||
// Each position represents a portion of the level
|
||||
threshold := (i + 1) * 100 / width
|
||||
if level >= threshold {
|
||||
meter += string(blocks[8]) // Full
|
||||
} else if level >= threshold-10 {
|
||||
// Partial - calculate which block to use
|
||||
partial := (level - (threshold - 10)) * 8 / 10
|
||||
if partial < 0 {
|
||||
partial = 0
|
||||
}
|
||||
meter += string(blocks[partial])
|
||||
} else {
|
||||
meter += string(blocks[0]) // Empty
|
||||
}
|
||||
}
|
||||
|
||||
return meter
|
||||
}
|
||||
286
pkg/audio/playback.go
Normal file
286
pkg/audio/playback.go
Normal file
@@ -0,0 +1,286 @@
|
||||
package audio
|
||||
|
||||
import (
|
||||
"encoding/binary"
|
||||
"fmt"
|
||||
"sync"
|
||||
"time"
|
||||
"unsafe"
|
||||
|
||||
"github.com/go-ole/go-ole"
|
||||
"github.com/moutend/go-wca/pkg/wca"
|
||||
)
|
||||
|
||||
// Player handles WASAPI audio playback
|
||||
type Player struct {
|
||||
client *wca.IAudioClient
|
||||
renderClient *wca.IAudioRenderClient
|
||||
waveFormat *wca.WAVEFORMATEX
|
||||
bufferSize uint32
|
||||
volume float32
|
||||
muted bool
|
||||
mu sync.Mutex
|
||||
running bool
|
||||
stopChan chan struct{}
|
||||
|
||||
// Audio buffer - accumulates incoming audio
|
||||
audioBuffer []int16
|
||||
bufferMu sync.Mutex
|
||||
|
||||
// Frame queue (960 samples = 20ms at 48kHz)
|
||||
frameQueue chan []int16
|
||||
}
|
||||
|
||||
const frameSamples = 960 // 20ms at 48kHz
|
||||
|
||||
// NewPlayer creates a new WASAPI audio player
|
||||
func NewPlayer() (*Player, error) {
|
||||
// Initialize COM using go-ole
|
||||
if err := ole.CoInitializeEx(0, ole.COINIT_APARTMENTTHREADED); err != nil {
|
||||
// Ignore if already initialized
|
||||
}
|
||||
|
||||
// Get default audio endpoint
|
||||
var deviceEnumerator *wca.IMMDeviceEnumerator
|
||||
if err := wca.CoCreateInstance(
|
||||
wca.CLSID_MMDeviceEnumerator,
|
||||
0,
|
||||
wca.CLSCTX_ALL,
|
||||
wca.IID_IMMDeviceEnumerator,
|
||||
&deviceEnumerator,
|
||||
); err != nil {
|
||||
return nil, fmt.Errorf("failed to create device enumerator: %w", err)
|
||||
}
|
||||
defer deviceEnumerator.Release()
|
||||
|
||||
var device *wca.IMMDevice
|
||||
if err := deviceEnumerator.GetDefaultAudioEndpoint(wca.ERender, wca.EConsole, &device); err != nil {
|
||||
return nil, fmt.Errorf("failed to get default render device: %w", err)
|
||||
}
|
||||
defer device.Release()
|
||||
|
||||
// Activate audio client
|
||||
var audioClient *wca.IAudioClient
|
||||
if err := device.Activate(wca.IID_IAudioClient, wca.CLSCTX_ALL, nil, &audioClient); err != nil {
|
||||
return nil, fmt.Errorf("failed to activate audio client: %w", err)
|
||||
}
|
||||
|
||||
// Set up format for 48kHz mono 16-bit (TeamSpeak format)
|
||||
waveFormat := &wca.WAVEFORMATEX{
|
||||
WFormatTag: wca.WAVE_FORMAT_PCM,
|
||||
NChannels: 1,
|
||||
NSamplesPerSec: 48000,
|
||||
WBitsPerSample: 16,
|
||||
NBlockAlign: 2,
|
||||
NAvgBytesPerSec: 96000,
|
||||
CbSize: 0,
|
||||
}
|
||||
|
||||
// Initialize in shared mode - 100ms buffer
|
||||
duration := wca.REFERENCE_TIME(100 * 10000) // 100ms in 100-nanosecond units
|
||||
if err := audioClient.Initialize(
|
||||
wca.AUDCLNT_SHAREMODE_SHARED,
|
||||
wca.AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM|wca.AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY,
|
||||
duration,
|
||||
0,
|
||||
waveFormat,
|
||||
nil,
|
||||
); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to initialize audio client: %w", err)
|
||||
}
|
||||
|
||||
// Get buffer size
|
||||
var bufferSize uint32
|
||||
if err := audioClient.GetBufferSize(&bufferSize); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to get buffer size: %w", err)
|
||||
}
|
||||
|
||||
// Get render client
|
||||
var renderClient *wca.IAudioRenderClient
|
||||
if err := audioClient.GetService(wca.IID_IAudioRenderClient, &renderClient); err != nil {
|
||||
audioClient.Release()
|
||||
return nil, fmt.Errorf("failed to get render client: %w", err)
|
||||
}
|
||||
|
||||
return &Player{
|
||||
client: audioClient,
|
||||
renderClient: renderClient,
|
||||
waveFormat: waveFormat,
|
||||
bufferSize: bufferSize,
|
||||
volume: 1.0,
|
||||
muted: false,
|
||||
stopChan: make(chan struct{}),
|
||||
audioBuffer: make([]int16, 0, frameSamples*50), // ~1 second buffer
|
||||
frameQueue: make(chan []int16, 100), // ~2 seconds of frames
|
||||
}, nil
|
||||
}
|
||||
|
||||
// Start begins audio playback
|
||||
func (p *Player) Start() error {
|
||||
p.mu.Lock()
|
||||
if p.running {
|
||||
p.mu.Unlock()
|
||||
return nil
|
||||
}
|
||||
p.running = true
|
||||
p.mu.Unlock()
|
||||
|
||||
if err := p.client.Start(); err != nil {
|
||||
return fmt.Errorf("failed to start audio client: %w", err)
|
||||
}
|
||||
|
||||
// Playback loop writes frames from queue to WASAPI
|
||||
go p.playbackLoop()
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// Stop stops audio playback
|
||||
func (p *Player) Stop() {
|
||||
p.mu.Lock()
|
||||
if !p.running {
|
||||
p.mu.Unlock()
|
||||
return
|
||||
}
|
||||
p.running = false
|
||||
p.mu.Unlock()
|
||||
|
||||
close(p.stopChan)
|
||||
p.client.Stop()
|
||||
}
|
||||
|
||||
// Close releases all resources
|
||||
func (p *Player) Close() {
|
||||
p.Stop()
|
||||
if p.renderClient != nil {
|
||||
p.renderClient.Release()
|
||||
}
|
||||
if p.client != nil {
|
||||
p.client.Release()
|
||||
}
|
||||
ole.CoUninitialize()
|
||||
}
|
||||
|
||||
// PlayPCM queues PCM audio for playback
|
||||
// Accumulates samples and queues complete 960-sample frames
|
||||
func (p *Player) PlayPCM(samples []int16) {
|
||||
if p.muted {
|
||||
return
|
||||
}
|
||||
|
||||
// Apply volume
|
||||
adjusted := samples
|
||||
if p.volume != 1.0 {
|
||||
adjusted = make([]int16, len(samples))
|
||||
for i, s := range samples {
|
||||
adjusted[i] = int16(float32(s) * p.volume)
|
||||
}
|
||||
}
|
||||
|
||||
p.bufferMu.Lock()
|
||||
p.audioBuffer = append(p.audioBuffer, adjusted...)
|
||||
|
||||
// Queue complete 960-sample frames
|
||||
for len(p.audioBuffer) >= frameSamples {
|
||||
frame := make([]int16, frameSamples)
|
||||
copy(frame, p.audioBuffer[:frameSamples])
|
||||
p.audioBuffer = p.audioBuffer[frameSamples:]
|
||||
|
||||
select {
|
||||
case p.frameQueue <- frame:
|
||||
default:
|
||||
// Queue full, drop oldest frame
|
||||
select {
|
||||
case <-p.frameQueue:
|
||||
default:
|
||||
}
|
||||
p.frameQueue <- frame
|
||||
}
|
||||
}
|
||||
p.bufferMu.Unlock()
|
||||
}
|
||||
|
||||
// SetVolume sets playback volume (0.0 to 1.0)
|
||||
func (p *Player) SetVolume(vol float32) {
|
||||
if vol < 0 {
|
||||
vol = 0
|
||||
}
|
||||
if vol > 1.0 {
|
||||
vol = 1.0
|
||||
}
|
||||
p.mu.Lock()
|
||||
p.volume = vol
|
||||
p.mu.Unlock()
|
||||
}
|
||||
|
||||
// GetVolume returns current volume (0.0 to 1.0)
|
||||
func (p *Player) GetVolume() float32 {
|
||||
p.mu.Lock()
|
||||
defer p.mu.Unlock()
|
||||
return p.volume
|
||||
}
|
||||
|
||||
// SetMuted sets mute state
|
||||
func (p *Player) SetMuted(muted bool) {
|
||||
p.mu.Lock()
|
||||
p.muted = muted
|
||||
p.mu.Unlock()
|
||||
}
|
||||
|
||||
// IsMuted returns mute state
|
||||
func (p *Player) IsMuted() bool {
|
||||
p.mu.Lock()
|
||||
defer p.mu.Unlock()
|
||||
return p.muted
|
||||
}
|
||||
|
||||
func (p *Player) playbackLoop() {
|
||||
// Use 20ms ticker matching TeamSpeak frame rate
|
||||
ticker := time.NewTicker(20 * time.Millisecond)
|
||||
defer ticker.Stop()
|
||||
|
||||
for {
|
||||
select {
|
||||
case <-p.stopChan:
|
||||
return
|
||||
case <-ticker.C:
|
||||
p.writeFrame()
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (p *Player) writeFrame() {
|
||||
// Get current padding (samples already in buffer)
|
||||
var padding uint32
|
||||
if err := p.client.GetCurrentPadding(&padding); err != nil {
|
||||
return
|
||||
}
|
||||
|
||||
available := p.bufferSize - padding
|
||||
if available < frameSamples {
|
||||
return // Not enough space for a full frame
|
||||
}
|
||||
|
||||
// Try to get a frame from the queue
|
||||
select {
|
||||
case frame := <-p.frameQueue:
|
||||
var buffer *byte
|
||||
if err := p.renderClient.GetBuffer(frameSamples, &buffer); err != nil {
|
||||
return
|
||||
}
|
||||
|
||||
// Write frame to WASAPI buffer
|
||||
bufSlice := unsafe.Slice(buffer, frameSamples*2)
|
||||
for i := 0; i < frameSamples; i++ {
|
||||
binary.LittleEndian.PutUint16(bufSlice[i*2:], uint16(frame[i]))
|
||||
}
|
||||
|
||||
p.renderClient.ReleaseBuffer(frameSamples, 0)
|
||||
|
||||
default:
|
||||
// No audio available - optionally write silence
|
||||
// (skip for now to avoid crackling)
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user