git.sesse.net Git - casparcg/blob - core/mixer/audio/audio_mixer.cpp

   1 /*
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>
   3 *
   4 * This file is part of CasparCG (www.casparcg.com).
   5 *
   6 * CasparCG is free software: you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License as published by
   8 * the Free Software Foundation, either version 3 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * CasparCG is distributed in the hope that it will be useful,
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 * GNU General Public License for more details.
  15 *
  16 * You should have received a copy of the GNU General Public License
  17 * along with CasparCG. If not, see <http://www.gnu.org/licenses/>.
  18 *
  19 * Author: Robert Nagy, ronag89@gmail.com
  20 */
  21
  22 #include "../../stdafx.h"
  23
  24 #include "audio_mixer.h"
  25
  26 #include <core/frame/frame.h>
  27 #include <core/frame/frame_transform.h>
  28 #include <common/diagnostics/graph.h>
  29
  30 #include <boost/range/adaptors.hpp>
  31 #include <boost/range/distance.hpp>
  32
  33 #include <map>
  34 #include <stack>
  35 #include <vector>
  36
  37 namespace caspar { namespace core {
  38
  39 struct audio_item
  40 {
  41         const void*                     tag;
  42         audio_transform         transform;
  43         audio_buffer            audio_data;
  44
  45         audio_item()
  46         {
  47         }
  48
  49         audio_item(audio_item&& other)
  50                 : tag(std::move(other.tag))
  51                 , transform(std::move(other.transform))
  52                 , audio_data(std::move(other.audio_data))
  53         {
  54         }
  55 };
  56
  57 typedef std::vector<float, tbb::cache_aligned_allocator<float>> audio_buffer_ps;
  58
  59 struct audio_stream
  60 {
  61         audio_transform         prev_transform;
  62         audio_buffer_ps         audio_data;
  63         bool                            is_still                = false;
  64 };
  65
  66 struct audio_mixer::impl : boost::noncopyable
  67 {
  68         std::stack<core::audio_transform>       transform_stack_;
  69         std::map<const void*, audio_stream>     audio_streams_;
  70         std::vector<audio_item>                         items_;
  71         std::vector<int>                                        audio_cadence_;
  72         video_format_desc                                       format_desc_;
  73         float                                                           master_volume_                  = 1.0f;
  74         float                                                           previous_master_volume_ = master_volume_;
  75 public:
  76         impl()
  77         {
  78                 transform_stack_.push(core::audio_transform());
  79         }
  80
  81         void push(const frame_transform& transform)
  82         {
  83                 transform_stack_.push(transform_stack_.top()*transform.audio_transform);
  84         }
  85
  86         void visit(const const_frame& frame)
  87         {
  88                 if(transform_stack_.top().volume < 0.002 || frame.audio_data().empty())
  89                         return;
  90
  91                 audio_item item;
  92                 item.tag                = frame.stream_tag();
  93                 item.transform  = transform_stack_.top();
  94                 item.audio_data = frame.audio_data();
  95
  96                 if(item.transform.is_still)
  97                         item.transform.volume = 0.0;
  98
  99                 items_.push_back(std::move(item));
 100         }
 101
 102         void begin(const core::audio_transform& transform)
 103         {
 104                 transform_stack_.push(transform_stack_.top()*transform);
 105         }
 106
 107         void pop()
 108         {
 109                 transform_stack_.pop();
 110         }
 111
 112         void set_master_volume(float volume)
 113         {
 114                 master_volume_ = volume;
 115         }
 116
 117         audio_buffer mix(const video_format_desc& format_desc)
 118         {
 119                 if(format_desc_ != format_desc)
 120                 {
 121                         audio_streams_.clear();
 122                         audio_cadence_ = format_desc.audio_cadence;
 123                         format_desc_ = format_desc;
 124                 }
 125
 126                 std::map<const void*, audio_stream>     next_audio_streams;
 127                 std::vector<const void*> used_tags;
 128
 129                 for (auto& item : items_)
 130                 {
 131                         audio_buffer_ps next_audio;
 132
 133                         auto next_transform = item.transform;
 134                         auto prev_transform = next_transform;
 135
 136                         auto tag = item.tag;
 137
 138                         if(boost::range::find(used_tags, tag) != used_tags.end())
 139                                 continue;
 140
 141                         used_tags.push_back(tag);
 142
 143                         const auto it = audio_streams_.find(tag);
 144                         if(it != audio_streams_.end())
 145                         {
 146                                 prev_transform  = it->second.prev_transform;
 147                                 next_audio              = std::move(it->second.audio_data);
 148                         }
 149
 150                         // Skip it if there is no existing audio stream and item has no audio-data.
 151                         if(it == audio_streams_.end() && item.audio_data.empty())
 152                                 continue;
 153
 154                         const float prev_volume = static_cast<float>(prev_transform.volume) * previous_master_volume_;
 155                         const float next_volume = static_cast<float>(next_transform.volume) * master_volume_;
 156
 157                         // TODO: Move volume mixing into code below, in order to support audio sample counts not corresponding to frame audio samples.
 158                         auto alpha = (next_volume-prev_volume)/static_cast<float>(item.audio_data.size()/format_desc.audio_channels);
 159
 160                         for(size_t n = 0; n < item.audio_data.size(); ++n)
 161                         {
 162                                 auto sample_multiplier = (prev_volume + (n/format_desc_.audio_channels) * alpha);
 163                                 next_audio.push_back(item.audio_data[n] * sample_multiplier);
 164                         }
 165
 166                         next_audio_streams[tag].prev_transform  = std::move(next_transform); // Store all active tags, inactive tags will be removed at the end.
 167                         next_audio_streams[tag].audio_data              = std::move(next_audio);
 168                         next_audio_streams[tag].is_still                = item.transform.is_still;
 169                 }
 170
 171                 previous_master_volume_ = master_volume_;
 172                 items_.clear();
 173
 174                 audio_streams_ = std::move(next_audio_streams);
 175
 176                 if(audio_streams_.empty())
 177                         audio_streams_[nullptr].audio_data = audio_buffer_ps(audio_size(audio_cadence_.front()), 0.0f);
 178
 179                 { // sanity check
 180
 181                         auto nb_invalid_streams = boost::count_if(audio_streams_ | boost::adaptors::map_values, [&](const audio_stream& x)
 182                         {
 183                                 return x.audio_data.size() < audio_size(audio_cadence_.front());
 184                         });
 185
 186                         if(nb_invalid_streams > 0)
 187                                 CASPAR_LOG(trace) << "[audio_mixer] Incorrect frame audio cadence detected.";
 188                 }
 189
 190                 std::vector<float> result_ps(audio_size(audio_cadence_.front()), 0.0f);
 191                 for (auto& stream : audio_streams_ | boost::adaptors::map_values)
 192                 {
 193                         if(stream.audio_data.size() < result_ps.size())
 194                                 stream.audio_data.resize(result_ps.size(), 0.0f);
 195
 196                         auto out = boost::range::transform(result_ps, stream.audio_data, std::begin(result_ps), std::plus<float>());
 197                         stream.audio_data.erase(std::begin(stream.audio_data), std::begin(stream.audio_data) + std::distance(std::begin(result_ps), out));
 198                 }
 199
 200                 boost::range::rotate(audio_cadence_, std::begin(audio_cadence_)+1);
 201
 202                 audio_buffer result;
 203                 result.reserve(result_ps.size());
 204                 boost::range::transform(result_ps, std::back_inserter(result), [](float sample){return static_cast<int32_t>(sample);});
 205
 206                 return result;
 207         }
 208
 209         size_t audio_size(size_t num_samples) const
 210         {
 211                 return num_samples * format_desc_.audio_channels;
 212         }
 213 };
 214
 215 audio_mixer::audio_mixer() : impl_(new impl()){}
 216 void audio_mixer::push(const frame_transform& transform){impl_->push(transform);}
 217 void audio_mixer::visit(const const_frame& frame){impl_->visit(frame);}
 218 void audio_mixer::pop(){impl_->pop();}
 219 void audio_mixer::set_master_volume(float volume) { impl_->set_master_volume(volume); }
 220 audio_buffer audio_mixer::operator()(const video_format_desc& format_desc){return impl_->mix(format_desc);}
 221
 222 }}