voice.cpp

hl2 source code. Do not use it illegal.

/*==============================================================================
   Copyright (c) Valve, LLC. 1999
   All Rights Reserved.  Proprietary.
--------------------------------------------------------------------------------
   Author : DSpeyrer
   Purpose: 
--------------------------------------------------------------------------------
 * $Log:                      $
 * 
 * $NoKeywords:
==============================================================================*/

#include <windows.h>
#include <math.h>
#include <mmsystem.h>
#include <stdio.h>
#include "tier0/dbg.h"
#include "convar.h"
#include "CircularBuffer.h"
#include "voice.h"
#include "voice_wavefile.h"
#include "voice_sound_engine_interface.h"
#include "sound.h"
#include "r_efx.h"
#include "cdll_int.h"
#include "voice_gain.h"
#include "voice_mixer_controls.h"

#include "SoundService.h"
#include "ivoicerecord.h"
#include "ivoicecodec.h"

#include "filesystem.h"
#include "../../FileSystem_Engine.h"


void Voice_EndChannel( int iChannel );


// Special entity index used for tweak mode.
#define TWEAKMODE_ENTITYINDEX				-500

// Special channel index passed to Voice_AddIncomingData when in tweak mode.
#define TWEAKMODE_CHANNELINDEX				-100


// How long does the sign stay above someone's head when they talk?
#define SPARK_TIME		0.2

// How long a voice channel has to be inactive before we free it.
#define DIE_COUNTDOWN	0.5

#define VOICE_RECEIVE_BUFFER_SIZE			(VOICE_OUTPUT_SAMPLE_RATE * BYTES_PER_SAMPLE)

#define TIME_PADDING						0.2f	// Time between receiving the first voice packet and actually starting
													// to play the sound. This accounts for frametime differences on the clients
													// and the server.

#define LOCALPLAYERTALKING_TIMEOUT			0.2f	// How long it takes for the client to decide the server isn't sending acks
													// of voice data back.


// The format we sample voice in.
WAVEFORMATEX g_VoiceSampleFormat =
{
	WAVE_FORMAT_PCM,		// wFormatTag
	1,						// nChannels
	8000,					// nSamplesPerSec
	16000,					// nAvgBytesPerSec
	2,						// nBlockAlign
	16,						// wBitsPerSample
	sizeof(WAVEFORMATEX)	// cbSize
};


ConVar voice_dsound( "voice_dsound", "0" ); 
ConVar voice_avggain( "voice_avggain", "0.5" );
ConVar voice_maxgain( "voice_maxgain", "5" );
ConVar voice_scale( "voice_scale", "1", FCVAR_ARCHIVE );

ConVar voice_loopback( "voice_loopback", "0" );
ConVar voice_fadeouttime( "voice_fadeouttime", "0.1" );	// It fades to no sound at the tail end of your voice data when you release the key.

// Debugging cvars.
ConVar voice_profile( "voice_profile", "0" );
ConVar voice_showchannels( "voice_showchannels", "0" );	// 1 = list channels
															// 2 = show timing info, etc
ConVar voice_showincoming( "voice_showincoming", "0" );	// show incoming voice data

ConVar voice_enable( "voice_enable", "1", FCVAR_ARCHIVE );		// Globally enable or disable voice.

// Have it force your mixer control settings so waveIn comes from the microphone. 
// CD rippers change your waveIn to come from the CD drive
ConVar voice_forcemicrecord( "voice_forcemicrecord", "1", FCVAR_ARCHIVE );

int g_nVoiceFadeSamples		= 1;							// Calculated each frame from the cvar.
float g_VoiceFadeMul		= 1;							// 1 / (g_nVoiceFadeSamples - 1).

// While in tweak mode, you can't hear anything anyone else is saying, and your own voice data
// goes directly to the speakers.
bool g_bInTweakMode = false;

bool g_bVoiceAtLeastPartiallyInitted = false;


IMixerControls *g_pMixerControls = NULL;


// Timing info for each frame.
static double	g_CompressTime = 0;
static double	g_DecompressTime = 0;
static double	g_GainTime = 0;
static double	g_UpsampleTime = 0;

class CVoiceTimer
{
public:
	inline void		Start()
	{
		if( voice_profile.GetInt() )
			m_StartTime = g_pSoundServices->GetRealTime();
	}
	
	inline void		End(double *out)
	{
		if( voice_profile.GetInt() )
			*out += g_pSoundServices->GetRealTime() - m_StartTime;
	}

	double			m_StartTime;
};


static bool		g_bLocalPlayerTalkingAck = false;
static float	g_LocalPlayerTalkingTimeout = 0;


CSysModule *g_hVoiceCodecDLL = 0;

// Voice recorder. Can be waveIn, DSound, or whatever.
static IVoiceRecord *g_pVoiceRecord = NULL;
static IVoiceCodec  *g_pEncodeCodec = NULL;

static bool			g_bVoiceRecording = false;	// Are we recording at the moment?

// Hacked functions to create the inputs and codecs..
extern IVoiceRecord*	CreateVoiceRecord_DSound(int nSamplesPerSec);

//
// Used for storing incoming voice data from an entity.
//
class CVoiceChannel
{
public:
									CVoiceChannel();

	// Called when someone speaks and a new voice channel is setup to hold the data.
	void							Init(int nEntity);

public:
	int								m_iEntity;		// Number of the entity speaking on this channel (index into cl_entities).
													// This is -1 when the channel is unused.

	
	CSizedCircularBuffer
		<VOICE_RECEIVE_BUFFER_SIZE>	m_Buffer;		// Circular buffer containing the voice data.

	// Used for upsampling..
	double							m_LastFraction;	// Fraction between m_LastSample and the next sample.
	short							m_LastSample;
	
	bool							m_bStarved;		// Set to true when the channel runs out of data for the mixer.
													// The channel is killed at that point.

	float							m_TimePad;		// Set to TIME_PADDING when the first voice packet comes in from a sender.
													// We add time padding (for frametime differences)
													// by waiting a certain amount of time before starting to output the sound.

	IVoiceCodec						*m_pVoiceCodec;	// Each channel gets is own IVoiceCodec instance so the codec can maintain state.

	CAutoGain						m_AutoGain;		// Gain we're applying to this channel

	CVoiceChannel					*m_pNext;
};


CVoiceChannel::CVoiceChannel()
{
	m_iEntity = -1;
	m_pVoiceCodec = NULL;
}

void CVoiceChannel::Init(int nEntity)
{
	m_iEntity = nEntity;
	m_bStarved = false;
	m_Buffer.Flush();
	m_TimePad = TIME_PADDING;
	m_LastSample = 0;
	m_LastFraction = 0.999;
	
	m_AutoGain.Reset( 128, voice_maxgain.GetFloat(), voice_avggain.GetFloat(), voice_scale.GetFloat() );
}



// Incoming voice channels.
CVoiceChannel g_VoiceChannels[VOICE_NUM_CHANNELS];


// These are used for recording the wave data into files for debugging.
#define MAX_WAVEFILEDATA_LEN	1024*1024
char *g_pUncompressedFileData = NULL;
int g_nUncompressedDataBytes = 0;
const char *g_pUncompressedDataFilename = NULL;

char *g_pDecompressedFileData = NULL;
int g_nDecompressedDataBytes = 0;
const char *g_pDecompressedDataFilename = NULL;

char *g_pMicInputFileData = NULL;
int g_nMicInputFileBytes = 0;
int g_CurMicInputFileByte = 0;
double g_MicStartTime;


inline void ApplyFadeToSamples(short *pSamples, int nSamples, int fadeOffset, float fadeMul)
{
	for(int i=0; i < nSamples; i++)
	{
		float percent = (i+fadeOffset) * fadeMul;
		pSamples[i] = (short)(pSamples[i] * (1 - percent));
	}
}


int Voice_GetOutputData(
	const int iChannel,			//! The voice channel it wants samples from.
	char *copyBufBytes,			//! The buffer to copy the samples into.
	const int copyBufSize,		//! Maximum size of copyBuf.
	const int samplePosition,	//! Which sample to start at.
	const int sampleCount		//! How many samples to get.
)
{
	CVoiceChannel *pChannel = &g_VoiceChannels[iChannel];	
	short *pCopyBuf = (short*)copyBufBytes;


	int maxOutSamples = copyBufSize / BYTES_PER_SAMPLE;

	// Find out how much we want and get it from the received data channel.	
	CCircularBuffer *pBuffer = &pChannel->m_Buffer;
	int nBytesToRead = pBuffer->GetReadAvailable();
	nBytesToRead = min(min(nBytesToRead, (int)maxOutSamples), sampleCount * BYTES_PER_SAMPLE);
	int nSamplesGotten = pBuffer->Read(pCopyBuf, nBytesToRead) / BYTES_PER_SAMPLE;

	// Are we at the end of the buffer's data? If so, fade data to silence so it doesn't clip.
	int readSamplesAvail = pBuffer->GetReadAvailable() / BYTES_PER_SAMPLE;
	if(readSamplesAvail < g_nVoiceFadeSamples)
	{
		int bufferFadeOffset = max((readSamplesAvail + nSamplesGotten) - g_nVoiceFadeSamples, 0);
		int globalFadeOffset = max(g_nVoiceFadeSamples - (readSamplesAvail + nSamplesGotten), 0);
		
		ApplyFadeToSamples(
			&pCopyBuf[bufferFadeOffset], 
			nSamplesGotten - bufferFadeOffset,
			globalFadeOffset,
			g_VoiceFadeMul);
	}
	
	// If there weren't enough bytes in the received data channel, pad it with zeros.
	if(nSamplesGotten < sampleCount)
	{
		int wantedSampleCount = min(sampleCount, maxOutSamples);
		memset(&pCopyBuf[nSamplesGotten], 0, (wantedSampleCount - nSamplesGotten) * BYTES_PER_SAMPLE);
		nSamplesGotten = wantedSampleCount;
	}

	// If the buffer is out of data, mark this channel to go away.
	if(pBuffer->GetReadAvailable() == 0)
	{
		pChannel->m_bStarved = true;
	}

	if(voice_showchannels.GetInt() >= 2)
	{
		Msg("Voice - mixed %d samples from channel %d\n", nSamplesGotten, iChannel);
	}

	VoiceSE_MoveMouth(pChannel->m_iEntity, (short*)copyBufBytes, nSamplesGotten);
	return nSamplesGotten;
}


void Voice_OnAudioSourceShutdown( int iChannel )
{
	Voice_EndChannel( iChannel );
}


// ------------------------------------------------------------------------ //
// Internal stuff.
// ------------------------------------------------------------------------ //

CVoiceChannel* GetVoiceChannel(int iChannel, bool bAssert=true)
{
	if(iChannel < 0 || iChannel >= VOICE_NUM_CHANNELS)
	{
		if(bAssert)
		{
			assert(false);
		}
		return NULL;
	}
	else
	{
		return &g_VoiceChannels[iChannel];
	}
}


bool Voice_Init(const char *pCodecName)
{
	if ( voice_enable.GetInt() == 0 )
	{
		return false;
	}

	Voice_Deinit();

	g_bVoiceAtLeastPartiallyInitted = true;

	if(!VoiceSE_Init())
		return false;

	// Get the voice input device.
	int rate = g_VoiceSampleFormat.nSamplesPerSec;
	g_pVoiceRecord = CreateVoiceRecord_DSound( rate );
	if( !g_pVoiceRecord )
	{
		Msg( "Unable to initialize DirectSoundCapture. You won't be able to speak to other players." );
	}

	// Get the codec.
	char dllName[256];
	Q_snprintf( dllName, sizeof( dllName ), "%s.dll", pCodecName );
	CreateInterfaceFn createCodecFn;
	g_hVoiceCodecDLL = FileSystem_LoadModule(dllName);

	if ( !g_hVoiceCodecDLL || !(createCodecFn = Sys_GetFactory(g_hVoiceCodecDLL)) ||
		 !(g_pEncodeCodec = (IVoiceCodec*)createCodecFn(pCodecName, NULL)) || !g_pEncodeCodec->Init() )
	{
		Msg("Unable to load voice codec '%s'. Voice disabled.\n", pCodecName);
		Voice_Deinit();
		return false;
	}

	for(int i=0; i < VOICE_NUM_CHANNELS; i++)
	{
		CVoiceChannel *pChannel = &g_VoiceChannels[i];

		if(!(pChannel->m_pVoiceCodec = (IVoiceCodec*)createCodecFn(pCodecName, NULL)) || !pChannel->m_pVoiceCodec->Init())
		{
			Voice_Deinit();
			return false;
		}
	}

	g_pMixerControls = GetMixerControls();
	
	if( voice_forcemicrecord.GetInt() )
	{
		if( g_pMixerControls )
			g_pMixerControls->SelectMicrophoneForWaveInput();
	}

	return true;
}


void Voice_EndChannel(int iChannel)
{
	assert(iChannel >= 0 && iChannel < VOICE_NUM_CHANNELS);

	CVoiceChannel *pChannel = &g_VoiceChannels[iChannel];
	
	if ( pChannel->m_iEntity != -1 )
	{
		int iEnt = pChannel->m_iEntity;
		pChannel->m_iEntity = -1;

		VoiceSE_EndChannel( iChannel );
		g_pSoundServices->OnChangeVoiceStatus( iEnt, false );
		VoiceSE_CloseMouth( iEnt );
	}
}


void Voice_EndAllChannels()
{
	for(int i=0; i < VOICE_NUM_CHANNELS; i++)
	{
		Voice_EndChannel(i);
	}
}


void Voice_Deinit()
{
	// This call tends to be expensive and when voice is not enabled it will continually
	// call in here, so avoid the work if possible.
	if( !g_bVoiceAtLeastPartiallyInitted )
		return;

	Voice_EndAllChannels();

	Voice_RecordStop();

	for(int i=0; i < VOICE_NUM_CHANNELS; i++)
	{
		CVoiceChannel *pChannel = &g_VoiceChannels[i];
		
		if(pChannel->m_pVoiceCodec)
		{
			pChannel->m_pVoiceCodec->Release();
			pChannel->m_pVoiceCodec = NULL;
		}
	}

	if(g_pEncodeCodec)
	{
		g_pEncodeCodec->Release();
		g_pEncodeCodec = NULL;
	}

	if(g_hVoiceCodecDLL)
	{
		FileSystem_UnloadModule(g_hVoiceCodecDLL);
		g_hVoiceCodecDLL = NULL;
	}

	if(g_pVoiceRecord)
	{
		g_pVoiceRecord->Release();
		g_pVoiceRecord = NULL;
	}

	VoiceSE_Term();

	if( g_pMixerControls )
	{
		g_pMixerControls->Release();
		g_pMixerControls = NULL;
	}

	g_bVoiceAtLeastPartiallyInitted = false;
}


bool Voice_GetLoopback()
{
	return !!voice_loopback.GetInt();
}


void Voice_LocalPlayerTalkingAck()
{
	if(!g_bLocalPlayerTalkingAck)
	{
		// Tell the client DLL when this changes.
		g_pSoundServices->OnChangeVoiceStatus(-2, TRUE);
	}

	g_bLocalPlayerTalkingAck = true;
	g_LocalPlayerTalkingTimeout = 0;
}


void Voice_UpdateVoiceTweakMode()
{
	if(!g_bInTweakMode || !g_pVoiceRecord)