2014-07-09 16:22:49 +02:00
|
|
|
// Copyright (c) 2014, Thomas Goyne <plorkyeran@aegisub.org>
|
|
|
|
//
|
|
|
|
// Permission to use, copy, modify, and distribute this software for any
|
|
|
|
// purpose with or without fee is hereby granted, provided that the above
|
|
|
|
// copyright notice and this permission notice appear in all copies.
|
|
|
|
//
|
|
|
|
// THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
|
|
|
|
// WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
|
|
|
|
// MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
|
|
|
|
// ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
|
|
|
// WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
|
|
|
|
// ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
|
|
|
|
// OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
|
|
|
|
//
|
|
|
|
// Aegisub Project http://www.aegisub.org/
|
|
|
|
|
|
|
|
#include "libaegisub/audio/provider.h"
|
|
|
|
|
|
|
|
#include "libaegisub/fs.h"
|
|
|
|
#include "libaegisub/io.h"
|
|
|
|
#include "libaegisub/log.h"
|
|
|
|
#include "libaegisub/util.h"
|
|
|
|
|
2022-08-10 15:09:41 +02:00
|
|
|
namespace {
|
|
|
|
|
|
|
|
template<typename Source>
|
|
|
|
class ConvertFloatToInt16 {
|
|
|
|
Source* src;
|
|
|
|
public:
|
|
|
|
ConvertFloatToInt16(Source* src) :src(src) {}
|
|
|
|
int16_t operator[](size_t idx) const {
|
|
|
|
Source expanded = src[idx] * 32768;
|
|
|
|
return expanded < -32768 ? -32768 :
|
|
|
|
expanded > 32767 ? 32767 :
|
|
|
|
static_cast<int16_t>(expanded);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
// 8 bits per sample is assumed to be unsigned with a bias of 128,
|
|
|
|
// while everything else is assumed to be signed with zero bias
|
|
|
|
class ConvertIntToInt16 {
|
|
|
|
void* src;
|
|
|
|
int bytes_per_sample;
|
|
|
|
public:
|
|
|
|
ConvertIntToInt16(void* src, int bytes_per_sample) :src(src), bytes_per_sample(bytes_per_sample) {}
|
|
|
|
const int16_t& operator[](size_t idx) const {
|
|
|
|
return *reinterpret_cast<int16_t*>(reinterpret_cast<char*>(src) + (idx + 1) * bytes_per_sample - sizeof(int16_t));
|
|
|
|
}
|
|
|
|
};
|
|
|
|
class ConvertUInt8ToInt16 {
|
|
|
|
uint8_t* src;
|
|
|
|
public:
|
|
|
|
ConvertUInt8ToInt16(uint8_t* src) :src(src) {}
|
|
|
|
int16_t operator[](size_t idx) const {
|
|
|
|
return int16_t(src[idx]-128) << 8;
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
template<typename Source>
|
|
|
|
class DownmixToMono {
|
|
|
|
Source src;
|
|
|
|
int channels;
|
|
|
|
public:
|
|
|
|
DownmixToMono(Source src, int channels) :src(src), channels(channels) {}
|
|
|
|
int16_t operator[](size_t idx) const {
|
|
|
|
int ret = 0;
|
|
|
|
// Just average the channels together
|
|
|
|
for (int i = 0; i < channels; ++i)
|
|
|
|
ret += src[idx * channels + i];
|
|
|
|
return ret / channels;
|
|
|
|
}
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2014-07-09 16:22:49 +02:00
|
|
|
namespace agi {
|
2022-08-10 15:09:41 +02:00
|
|
|
void AudioProvider::FillBufferInt16Mono(int16_t* buf, int64_t start, int64_t count) const {
|
|
|
|
if (!float_samples && bytes_per_sample == 2 && channels == 1) {
|
|
|
|
FillBuffer(buf, start, count);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
void* buff = malloc(bytes_per_sample * count * channels);
|
|
|
|
FillBuffer(buff, start, count);
|
|
|
|
if (channels == 1) {
|
|
|
|
if (float_samples) {
|
|
|
|
if (bytes_per_sample == sizeof(float))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = ConvertFloatToInt16<float>(reinterpret_cast<float*>(buff))[i];
|
|
|
|
else if (bytes_per_sample == sizeof(double))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = ConvertFloatToInt16<double>(reinterpret_cast<double*>(buff))[i];
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if (bytes_per_sample == sizeof(uint8_t))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = ConvertUInt8ToInt16(reinterpret_cast<uint8_t*>(buff))[i];
|
|
|
|
else
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = ConvertIntToInt16(buff, bytes_per_sample)[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if (float_samples) {
|
|
|
|
if (bytes_per_sample == sizeof(float))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = DownmixToMono<ConvertFloatToInt16<float> >(ConvertFloatToInt16<float>(reinterpret_cast<float*>(buff)), channels)[i];
|
|
|
|
else if (bytes_per_sample == sizeof(double))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = DownmixToMono<ConvertFloatToInt16<double> >(ConvertFloatToInt16<double>(reinterpret_cast<double*>(buff)), channels)[i];
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if (bytes_per_sample == sizeof(uint8_t))
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = DownmixToMono<ConvertUInt8ToInt16>(ConvertUInt8ToInt16(reinterpret_cast<uint8_t*>(buff)), channels)[i];
|
|
|
|
else
|
|
|
|
for (int64_t i = 0; i < count; ++i)
|
|
|
|
buf[i] = DownmixToMono<ConvertIntToInt16>(ConvertIntToInt16(buff, bytes_per_sample), channels)[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
free(buff);
|
|
|
|
}
|
2014-07-09 16:22:49 +02:00
|
|
|
|
2022-08-18 22:44:19 +02:00
|
|
|
// This entire file has turned into a mess. For now I'm just following the pattern of the wangqr code, but
|
|
|
|
// this should really be restructured entirely again. The original type constructor-based system worked very well - it could
|
|
|
|
// just give downmix/conversion control to the players instead.
|
|
|
|
void AudioProvider::GetAudioWithVolume(void *buf, int64_t start, int64_t count, double volume) const {
|
|
|
|
GetAudio(buf, start, count);
|
|
|
|
if (volume == 1.0) return;
|
|
|
|
|
|
|
|
int64_t n = count * GetChannels();
|
|
|
|
|
|
|
|
if (float_samples) {
|
|
|
|
if (bytes_per_sample == sizeof(float)) {
|
|
|
|
float *buff = reinterpret_cast<float *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = static_cast<float>(buff[i] * volume);
|
|
|
|
} else if (bytes_per_sample == sizeof(double)) {
|
|
|
|
double *buff = reinterpret_cast<double *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = buff[i] * volume;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
if (bytes_per_sample == sizeof(uint8_t)) {
|
|
|
|
uint8_t *buff = reinterpret_cast<uint8_t *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = util::mid(0, static_cast<int>(((int) buff[i] - 128) * volume + 128), 0xFF);
|
|
|
|
} else if (bytes_per_sample == sizeof(int16_t)) {
|
|
|
|
int16_t *buff = reinterpret_cast<int16_t *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = util::mid(-0x8000, static_cast<int>(buff[i] * volume), 0x7FFF);
|
|
|
|
} else if (bytes_per_sample == sizeof(int32_t)) {
|
|
|
|
int32_t *buff = reinterpret_cast<int32_t *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = static_cast<int32_t>(buff[i] * volume);
|
|
|
|
} else if (bytes_per_sample == sizeof(int64_t)) {
|
|
|
|
int64_t *buff = reinterpret_cast<int64_t *>(buf);
|
|
|
|
for (int64_t i = 0; i < n; ++i)
|
|
|
|
buff[i] = static_cast<int64_t>(buff[i] * volume);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-08-10 15:09:41 +02:00
|
|
|
void AudioProvider::GetInt16MonoAudioWithVolume(int16_t *buf, int64_t start, int64_t count, double volume) const {
|
|
|
|
GetInt16MonoAudio(buf, start, count);
|
2014-07-09 16:22:49 +02:00
|
|
|
if (volume == 1.0) return;
|
|
|
|
|
|
|
|
auto buffer = static_cast<int16_t *>(buf);
|
|
|
|
for (size_t i = 0; i < (size_t)count; ++i)
|
2014-07-14 17:55:47 +02:00
|
|
|
buffer[i] = util::mid(-0x8000, static_cast<int>(buffer[i] * volume + 0.5), 0x7FFF);
|
2014-07-09 16:22:49 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
void AudioProvider::ZeroFill(void *buf, int64_t count) const {
|
|
|
|
if (bytes_per_sample == 1)
|
2015-07-28 21:02:41 +02:00
|
|
|
// 8 bit formats are usually unsigned with bias 128
|
|
|
|
memset(buf, 128, count * channels);
|
2014-07-09 16:22:49 +02:00
|
|
|
else // While everything else is signed
|
|
|
|
memset(buf, 0, count * bytes_per_sample * channels);
|
|
|
|
}
|
|
|
|
|
|
|
|
void AudioProvider::GetAudio(void *buf, int64_t start, int64_t count) const {
|
|
|
|
if (start < 0) {
|
|
|
|
ZeroFill(buf, std::min(-start, count));
|
|
|
|
buf = static_cast<char *>(buf) + -start * bytes_per_sample * channels;
|
|
|
|
count += start;
|
|
|
|
start = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (start + count > num_samples) {
|
|
|
|
int64_t zero_count = std::min(count, start + count - num_samples);
|
|
|
|
count -= zero_count;
|
|
|
|
ZeroFill(static_cast<char *>(buf) + count * bytes_per_sample * channels, zero_count);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count <= 0) return;
|
|
|
|
|
|
|
|
try {
|
|
|
|
FillBuffer(buf, start, count);
|
|
|
|
}
|
|
|
|
catch (AudioDecodeError const& e) {
|
|
|
|
// We don't have any good way to report errors here, so just log the
|
|
|
|
// failure and return silence
|
|
|
|
LOG_E("audio_provider") << e.GetMessage();
|
|
|
|
ZeroFill(buf, count);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
catch (...) {
|
|
|
|
LOG_E("audio_provider") << "Unknown audio decoding error";
|
|
|
|
ZeroFill(buf, count);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-08-10 15:09:41 +02:00
|
|
|
void AudioProvider::GetInt16MonoAudio(int16_t* buf, int64_t start, int64_t count) const {
|
|
|
|
if (start < 0) {
|
|
|
|
memset(buf, 0, sizeof(int16_t) * std::min(-start, count));
|
|
|
|
buf -= start;
|
|
|
|
count += start;
|
|
|
|
start = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (start + count > num_samples) {
|
|
|
|
int64_t zero_count = std::min(count, start + count - num_samples);
|
|
|
|
count -= zero_count;
|
|
|
|
memset(buf + count, 0, sizeof(int16_t) * zero_count);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count <= 0) return;
|
|
|
|
|
|
|
|
try {
|
|
|
|
FillBufferInt16Mono(buf, start, count);
|
|
|
|
}
|
|
|
|
catch (AudioDecodeError const& e) {
|
|
|
|
// We don't have any good way to report errors here, so just log the
|
|
|
|
// failure and return silence
|
|
|
|
LOG_E("audio_provider") << e.GetMessage();
|
|
|
|
memset(buf, 0, sizeof(int16_t) * count);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
catch (...) {
|
|
|
|
LOG_E("audio_provider") << "Unknown audio decoding error";
|
|
|
|
memset(buf, 0, sizeof(int16_t) * count);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-07-09 16:22:49 +02:00
|
|
|
namespace {
|
|
|
|
class writer {
|
|
|
|
io::Save outfile;
|
|
|
|
std::ostream& out;
|
|
|
|
|
|
|
|
public:
|
|
|
|
writer(agi::fs::path const& filename) : outfile(filename, true), out(outfile.Get()) { }
|
|
|
|
|
|
|
|
template<int N>
|
|
|
|
void write(const char(&str)[N]) {
|
|
|
|
out.write(str, N - 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
void write(std::vector<char> const& data) {
|
|
|
|
out.write(data.data(), data.size());
|
|
|
|
}
|
|
|
|
|
|
|
|
template<typename Dest, typename Src>
|
|
|
|
void write(Src v) {
|
|
|
|
auto converted = static_cast<Dest>(v);
|
|
|
|
out.write(reinterpret_cast<char *>(&converted), sizeof(Dest));
|
|
|
|
}
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2015-07-28 21:02:41 +02:00
|
|
|
void SaveAudioClip(AudioProvider const& provider, fs::path const& path, int start_time, int end_time) {
|
|
|
|
const auto max_samples = provider.GetNumSamples();
|
|
|
|
const auto start_sample = std::min(max_samples, ((int64_t)start_time * provider.GetSampleRate() + 999) / 1000);
|
|
|
|
const auto end_sample = util::mid(start_sample, ((int64_t)end_time * provider.GetSampleRate() + 999) / 1000, max_samples);
|
2014-07-09 16:22:49 +02:00
|
|
|
|
2015-07-28 21:02:41 +02:00
|
|
|
const size_t bytes_per_sample = provider.GetBytesPerSample() * provider.GetChannels();
|
|
|
|
const size_t bufsize = (end_sample - start_sample) * bytes_per_sample;
|
2014-07-09 16:22:49 +02:00
|
|
|
|
|
|
|
writer out{path};
|
|
|
|
out.write("RIFF");
|
|
|
|
out.write<int32_t>(bufsize + 36);
|
|
|
|
|
|
|
|
out.write("WAVEfmt ");
|
|
|
|
out.write<int32_t>(16); // Size of chunk
|
2022-08-10 15:23:56 +02:00
|
|
|
out.write<int16_t>(provider.AreSamplesFloat() ? 3 : 1); // compression format (1: WAVE_FORMAT_PCM, 3: WAVE_FORMAT_IEEE_FLOAT)
|
2015-07-28 21:02:41 +02:00
|
|
|
out.write<int16_t>(provider.GetChannels());
|
|
|
|
out.write<int32_t>(provider.GetSampleRate());
|
|
|
|
out.write<int32_t>(provider.GetSampleRate() * provider.GetChannels() * provider.GetBytesPerSample());
|
|
|
|
out.write<int16_t>(provider.GetChannels() * provider.GetBytesPerSample());
|
|
|
|
out.write<int16_t>(provider.GetBytesPerSample() * 8);
|
2014-07-09 16:22:49 +02:00
|
|
|
|
|
|
|
out.write("data");
|
|
|
|
out.write<int32_t>(bufsize);
|
|
|
|
|
|
|
|
// samples per read
|
|
|
|
size_t spr = 65536 / bytes_per_sample;
|
|
|
|
std::vector<char> buf;
|
|
|
|
for (int64_t i = start_sample; i < end_sample; i += spr) {
|
|
|
|
spr = std::min<size_t>(spr, end_sample - i);
|
|
|
|
buf.resize(spr * bytes_per_sample);
|
2015-07-28 21:02:41 +02:00
|
|
|
provider.GetAudio(&buf[0], i, spr);
|
2014-07-09 16:22:49 +02:00
|
|
|
out.write(buf);
|
|
|
|
}
|
|
|
|
}
|
2022-08-18 22:44:19 +02:00
|
|
|
}
|