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 #include <common/linq.h>
  30
  31 #include <boost/range/adaptors.hpp>
  32 #include <boost/range/distance.hpp>
  33
  34 #include <map>
  35 #include <stack>
  36 #include <vector>
  37
  38 namespace caspar { namespace core {
  39
  40 struct audio_item
  41 {
  42         const void*                     tag                     = nullptr;
  43         audio_transform         transform;
  44         audio_buffer            audio_data;
  45
  46         audio_item()
  47         {
  48         }
  49
  50         audio_item(audio_item&& other)
  51                 : tag(std::move(other.tag))
  52                 , transform(std::move(other.transform))
  53                 , audio_data(std::move(other.audio_data))
  54         {
  55         }
  56 };
  57
  58 typedef cache_aligned_vector<float> audio_buffer_ps;
  59
  60 struct audio_stream
  61 {
  62         audio_transform         prev_transform;
  63         audio_buffer_ps         audio_data;
  64         bool                            is_still                = false;
  65 };
  66
  67 struct audio_mixer::impl : boost::noncopyable
  68 {
  69         std::stack<core::audio_transform>       transform_stack_;
  70         std::map<const void*, audio_stream>     audio_streams_;
  71         std::vector<audio_item>                         items_;
  72         std::vector<int>                                        audio_cadence_;
  73         video_format_desc                                       format_desc_;
  74         float                                                           master_volume_                  = 1.0f;
  75         float                                                           previous_master_volume_ = master_volume_;
  76 public:
  77         impl()
  78         {
  79                 transform_stack_.push(core::audio_transform());
  80         }
  81
  82         void push(const frame_transform& transform)
  83         {
  84                 transform_stack_.push(transform_stack_.top()*transform.audio_transform);
  85         }
  86
  87         void visit(const const_frame& frame)
  88         {
  89                 if(transform_stack_.top().volume < 0.002 || frame.audio_data().empty())
  90                         return;
  91
  92                 audio_item item;
  93                 item.tag                = frame.stream_tag();
  94                 item.transform  = transform_stack_.top();
  95                 item.audio_data = frame.audio_data();
  96
  97                 if(item.transform.is_still)
  98                         item.transform.volume = 0.0;
  99
 100                 items_.push_back(std::move(item));
 101         }
 102
 103         void begin(const core::audio_transform& transform)
 104         {
 105                 transform_stack_.push(transform_stack_.top()*transform);
 106         }
 107
 108         void pop()
 109         {
 110                 transform_stack_.pop();
 111         }
 112
 113         void set_master_volume(float volume)
 114         {
 115                 master_volume_ = volume;
 116         }
 117
 118         float get_master_volume()
 119         {
 120                 return master_volume_;
 121         }
 122
 123         audio_buffer mix(const video_format_desc& format_desc)
 124         {
 125                 if(format_desc_ != format_desc)
 126                 {
 127                         audio_streams_.clear();
 128                         audio_cadence_ = format_desc.audio_cadence;
 129                         format_desc_ = format_desc;
 130                 }
 131
 132                 std::map<const void*, audio_stream>     next_audio_streams;
 133                 std::vector<const void*> used_tags;
 134
 135                 for (auto& item : items_)
 136                 {
 137                         audio_buffer_ps next_audio;
 138
 139                         auto next_transform = item.transform;
 140                         auto prev_transform = next_transform;
 141
 142                         auto tag = item.tag;
 143
 144                         if(boost::range::find(used_tags, tag) != used_tags.end())
 145                                 continue;
 146
 147                         used_tags.push_back(tag);
 148
 149                         const auto it = audio_streams_.find(tag);
 150                         if(it != audio_streams_.end())
 151                         {
 152                                 prev_transform  = it->second.prev_transform;
 153                                 next_audio              = std::move(it->second.audio_data);
 154                         }
 155
 156                         // Skip it if there is no existing audio stream and item has no audio-data.
 157                         if(it == audio_streams_.end() && item.audio_data.empty())
 158                                 continue;
 159
 160                         const float prev_volume = static_cast<float>(prev_transform.volume) * previous_master_volume_;
 161                         const float next_volume = static_cast<float>(next_transform.volume) * master_volume_;
 162
 163                         // TODO: Move volume mixing into code below, in order to support audio sample counts not corresponding to frame audio samples.
 164                         auto alpha = (next_volume-prev_volume)/static_cast<float>(item.audio_data.size()/format_desc.audio_channels);
 165
 166                         for(size_t n = 0; n < item.audio_data.size(); ++n)
 167                         {
 168                                 auto sample_multiplier = (prev_volume + (n/format_desc_.audio_channels) * alpha);
 169                                 next_audio.push_back(item.audio_data[n] * sample_multiplier);
 170                         }
 171
 172                         next_audio_streams[tag].prev_transform  = std::move(next_transform); // Store all active tags, inactive tags will be removed at the end.
 173                         next_audio_streams[tag].audio_data              = std::move(next_audio);
 174                         next_audio_streams[tag].is_still                = item.transform.is_still;
 175                 }
 176
 177                 previous_master_volume_ = master_volume_;
 178                 items_.clear();
 179
 180                 audio_streams_ = std::move(next_audio_streams);
 181
 182                 if(audio_streams_.empty())
 183                         audio_streams_[nullptr].audio_data = audio_buffer_ps(audio_size(audio_cadence_.front()), 0.0f);
 184
 185                 { // sanity check
 186
 187                         auto nb_invalid_streams = cpplinq::from(audio_streams_)
 188                                 .select(values())
 189                                 .where([&](const audio_stream& x) { return x.audio_data.size() < audio_size(audio_cadence_.front()); })
 190                                 .count();
 191
 192                         if(nb_invalid_streams > 0)
 193                                 CASPAR_LOG(trace) << "[audio_mixer] Incorrect frame audio cadence detected.";
 194                 }
 195
 196                 std::vector<float> result_ps(audio_size(audio_cadence_.front()), 0.0f);
 197                 for (auto& stream : audio_streams_ | boost::adaptors::map_values)
 198                 {
 199                         if(stream.audio_data.size() < result_ps.size())
 200                                 stream.audio_data.resize(result_ps.size(), 0.0f);
 201
 202                         auto out = boost::range::transform(result_ps, stream.audio_data, std::begin(result_ps), std::plus<float>());
 203                         stream.audio_data.erase(std::begin(stream.audio_data), std::begin(stream.audio_data) + std::distance(std::begin(result_ps), out));
 204                 }
 205
 206                 boost::range::rotate(audio_cadence_, std::begin(audio_cadence_)+1);
 207
 208                 audio_buffer result;
 209                 result.reserve(result_ps.size());
 210                 boost::range::transform(result_ps, std::back_inserter(result), [](float sample){return static_cast<int32_t>(sample);});
 211
 212                 return result;
 213         }
 214
 215         size_t audio_size(size_t num_samples) const
 216         {
 217                 return num_samples * format_desc_.audio_channels;
 218         }
 219 };
 220
 221 audio_mixer::audio_mixer() : impl_(new impl()){}
 222 void audio_mixer::push(const frame_transform& transform){impl_->push(transform);}
 223 void audio_mixer::visit(const const_frame& frame){impl_->visit(frame);}
 224 void audio_mixer::pop(){impl_->pop();}
 225 void audio_mixer::set_master_volume(float volume) { impl_->set_master_volume(volume); }
 226 float audio_mixer::get_master_volume() { return impl_->get_master_volume(); }
 227 audio_buffer audio_mixer::operator()(const video_format_desc& format_desc){ return impl_->mix(format_desc); }
 228
 229 }}