#include <core/frame/frame.h>
#include <core/frame/frame_transform.h>
+#include <core/frame/audio_channel_layout.h>
#include <core/monitor/monitor.h>
#include <common/diagnostics/graph.h>
struct audio_item
{
- const void* tag = nullptr;
- audio_transform transform;
- audio_buffer audio_data;
+ const void* tag = nullptr;
+ audio_transform transform;
+ audio_buffer audio_data;
+ audio_channel_layout channel_layout = audio_channel_layout::invalid();
audio_item()
{
: tag(std::move(other.tag))
, transform(std::move(other.transform))
, audio_data(std::move(other.audio_data))
+ , channel_layout(std::move(other.channel_layout))
{
}
};
-typedef cache_aligned_vector<float> audio_buffer_ps;
-
+typedef cache_aligned_vector<double> audio_buffer_ps;
+
struct audio_stream
{
- audio_transform prev_transform;
- audio_buffer_ps audio_data;
- bool is_still = false;
+ audio_transform prev_transform;
+ audio_buffer_ps audio_data;
+ std::unique_ptr<audio_channel_remapper> channel_remapper;
+ bool remapping_failed = false;
+ bool is_still = false;
};
struct audio_mixer::impl : boost::noncopyable
std::vector<audio_item> items_;
std::vector<int> audio_cadence_;
video_format_desc format_desc_;
+ audio_channel_layout channel_layout_ = audio_channel_layout::invalid();
float master_volume_ = 1.0f;
float previous_master_volume_ = master_volume_;
spl::shared_ptr<diagnostics::graph> graph_;
: graph_(std::move(graph))
{
graph_->set_color("volume", diagnostics::color(1.0f, 0.8f, 0.1f));
+ graph_->set_color("audio-clipping", diagnostics::color(0.3f, 0.6f, 0.3f));
transform_stack_.push(core::audio_transform());
}
-
+
void push(const frame_transform& transform)
{
transform_stack_.push(transform_stack_.top()*transform.audio_transform);
return;
audio_item item;
- item.tag = frame.stream_tag();
- item.transform = transform_stack_.top();
- item.audio_data = frame.audio_data();
+ item.tag = frame.stream_tag();
+ item.transform = transform_stack_.top();
+ item.audio_data = frame.audio_data();
+ item.channel_layout = frame.audio_channel_layout();
if(item.transform.is_still)
item.transform.volume = 0.0;
-
- items_.push_back(std::move(item));
- }
- void begin(const core::audio_transform& transform)
- {
- transform_stack_.push(transform_stack_.top()*transform);
+ items_.push_back(std::move(item));
}
-
+
void pop()
{
transform_stack_.pop();
return master_volume_;
}
- audio_buffer mix(const video_format_desc& format_desc)
- {
- if(format_desc_ != format_desc)
+ audio_buffer mix(const video_format_desc& format_desc, const audio_channel_layout& channel_layout)
+ {
+ if(format_desc_ != format_desc || channel_layout_ != channel_layout)
{
audio_streams_.clear();
audio_cadence_ = format_desc.audio_cadence;
format_desc_ = format_desc;
- }
-
+ channel_layout_ = channel_layout;
+ }
+
std::map<const void*, audio_stream> next_audio_streams;
- std::vector<const void*> used_tags;
for (auto& item : items_)
- {
+ {
audio_buffer_ps next_audio;
+ std::unique_ptr<audio_channel_remapper> channel_remapper;
+ bool remapping_failed = false;
auto next_transform = item.transform;
auto prev_transform = next_transform;
auto tag = item.tag;
- if(boost::range::find(used_tags, tag) != used_tags.end())
+ auto it = next_audio_streams.find(tag);
+ bool found = it != next_audio_streams.end();
+
+ if (!found)
+ {
+ it = audio_streams_.find(tag);
+ found = it != audio_streams_.end();
+ }
+
+ if (found)
+ {
+ prev_transform = it->second.prev_transform;
+ next_audio = std::move(it->second.audio_data);
+ channel_remapper = std::move(it->second.channel_remapper);
+ remapping_failed = it->second.remapping_failed;
+ }
+
+ if (remapping_failed)
+ {
+ CASPAR_LOG(trace) << "[audio_mixer] audio channel remapping already failed for stream.";
+ next_audio_streams[tag].remapping_failed = true;
continue;
-
- used_tags.push_back(tag);
-
- const auto it = audio_streams_.find(tag);
- if(it != audio_streams_.end())
- {
- prev_transform = it->second.prev_transform;
- next_audio = std::move(it->second.audio_data);
}
-
+
// Skip it if there is no existing audio stream and item has no audio-data.
- if(it == audio_streams_.end() && item.audio_data.empty())
+ if(!found && item.audio_data.empty())
continue;
-
- const float prev_volume = static_cast<float>(prev_transform.volume) * previous_master_volume_;
- const float next_volume = static_cast<float>(next_transform.volume) * master_volume_;
+
+ if (item.channel_layout == audio_channel_layout::invalid())
+ {
+ CASPAR_LOG(warning) << "[audio_mixer] invalid audio channel layout for item";
+ next_audio_streams[tag].remapping_failed = true;
+ continue;
+ }
+
+ if (!channel_remapper)
+ {
+ try
+ {
+ channel_remapper.reset(new audio_channel_remapper(item.channel_layout, channel_layout_));
+ }
+ catch (...)
+ {
+ CASPAR_LOG_CURRENT_EXCEPTION();
+ CASPAR_LOG(error) << "[audio_mixer] audio channel remapping failed for stream.";
+ next_audio_streams[tag].remapping_failed = true;
+ continue;
+ }
+ }
+
+ item.audio_data = channel_remapper->mix_and_rearrange(item.audio_data);
+
+ const double prev_volume = prev_transform.volume * previous_master_volume_;
+ const double next_volume = next_transform.volume * master_volume_;
// TODO: Move volume mixing into code below, in order to support audio sample counts not corresponding to frame audio samples.
- auto alpha = (next_volume-prev_volume)/static_cast<float>(item.audio_data.size()/format_desc.audio_channels);
-
+ auto alpha = (next_volume-prev_volume)/static_cast<double>(item.audio_data.size()/channel_layout_.num_channels);
+
for(size_t n = 0; n < item.audio_data.size(); ++n)
{
- auto sample_multiplier = (prev_volume + (n/format_desc_.audio_channels) * alpha);
- next_audio.push_back(item.audio_data[n] * sample_multiplier);
- }
-
- next_audio_streams[tag].prev_transform = std::move(next_transform); // Store all active tags, inactive tags will be removed at the end.
- next_audio_streams[tag].audio_data = std::move(next_audio);
- next_audio_streams[tag].is_still = item.transform.is_still;
- }
+ auto sample_multiplier = (prev_volume + (n / channel_layout_.num_channels) * alpha);
+ next_audio.push_back(item.audio_data.data()[n] * sample_multiplier);
+ }
+
+ next_audio_streams[tag].prev_transform = std::move(next_transform); // Store all active tags, inactive tags will be removed at the end.
+ next_audio_streams[tag].audio_data = std::move(next_audio);
+ next_audio_streams[tag].channel_remapper = std::move(channel_remapper);
+ next_audio_streams[tag].remapping_failed = remapping_failed;
+ next_audio_streams[tag].is_still = item.transform.is_still;
+ }
previous_master_volume_ = master_volume_;
items_.clear();
audio_streams_ = std::move(next_audio_streams);
-
- if(audio_streams_.empty())
- audio_streams_[nullptr].audio_data = audio_buffer_ps(audio_size(audio_cadence_.front()), 0.0f);
+
+ if(audio_streams_.empty())
+ audio_streams_[nullptr].audio_data = audio_buffer_ps(audio_size(audio_cadence_.front()), 0.0);
{ // sanity check
auto nb_invalid_streams = cpplinq::from(audio_streams_)
.select(values())
- .where([&](const audio_stream& x) { return x.audio_data.size() < audio_size(audio_cadence_.front()); })
+ .where([&](const audio_stream& x)
+ {
+ return !x.remapping_failed && x.audio_data.size() < audio_size(audio_cadence_.front());
+ })
.count();
- if(nb_invalid_streams > 0)
- CASPAR_LOG(trace) << "[audio_mixer] Incorrect frame audio cadence detected.";
+ if(nb_invalid_streams > 0)
+ CASPAR_LOG(trace) << "[audio_mixer] Incorrect frame audio cadence detected.";
}
-
- std::vector<float> result_ps(audio_size(audio_cadence_.front()), 0.0f);
+
+ audio_buffer_ps result_ps(audio_size(audio_cadence_.front()), 0.0);
for (auto& stream : audio_streams_ | boost::adaptors::map_values)
{
- if(stream.audio_data.size() < result_ps.size())
- stream.audio_data.resize(result_ps.size(), 0.0f);
+ if (stream.audio_data.size() < result_ps.size())
+ {
+ auto samples = (result_ps.size() - stream.audio_data.size()) / channel_layout_.num_channels;
+ CASPAR_LOG(trace) << L"[audio_mixer] Appended " << samples << L" zero samples";
+ CASPAR_LOG(trace) << L"[audio_mixer] Actual number of samples " << stream.audio_data.size() / channel_layout_.num_channels;
+ CASPAR_LOG(trace) << L"[audio_mixer] Wanted number of samples " << result_ps.size() / channel_layout_.num_channels;
+ stream.audio_data.resize(result_ps.size(), 0.0);
+ }
- auto out = boost::range::transform(result_ps, stream.audio_data, std::begin(result_ps), std::plus<float>());
+ auto out = boost::range::transform(result_ps, stream.audio_data, std::begin(result_ps), std::plus<double>());
stream.audio_data.erase(std::begin(stream.audio_data), std::begin(stream.audio_data) + std::distance(std::begin(result_ps), out));
- }
-
+ }
+
boost::range::rotate(audio_cadence_, std::begin(audio_cadence_)+1);
-
- audio_buffer result;
+
+ auto result_owner = spl::make_shared<mutable_audio_buffer>();
+ auto& result = *result_owner;
result.reserve(result_ps.size());
- boost::range::transform(result_ps, std::back_inserter(result), [](float sample){return static_cast<int32_t>(sample);});
-
- const int num_channels = format_desc_.audio_channels;
+ const int32_t min_amplitude = std::numeric_limits<int32_t>::min();
+ const int32_t max_amplitude = std::numeric_limits<int32_t>::max();
+ bool clipping = false;
+ boost::range::transform(result_ps, std::back_inserter(result), [&](double sample)
+ {
+ if (sample > max_amplitude)
+ {
+ clipping = true;
+ return max_amplitude;
+ }
+ else if (sample < min_amplitude)
+ {
+ clipping = true;
+ return min_amplitude;
+ }
+ else
+ return static_cast<int32_t>(sample);
+ });
+
+ if (clipping)
+ graph_->set_tag(diagnostics::tag_severity::WARNING, "audio-clipping");
+
+ const int num_channels = channel_layout_.num_channels;
monitor_subject_ << monitor::message("/nb_channels") % num_channels;
auto max = std::vector<int32_t>(num_channels, std::numeric_limits<int32_t>::min());
graph_->set_value("volume", static_cast<double>(*boost::max_element(max)) / std::numeric_limits<int32_t>::max());
- return result;
+ return caspar::array<int32_t>(result.data(), result.size(), true, std::move(result_owner));
}
size_t audio_size(size_t num_samples) const
{
- return num_samples * format_desc_.audio_channels;
+ return num_samples * channel_layout_.num_channels;
}
};
void audio_mixer::pop(){impl_->pop();}
void audio_mixer::set_master_volume(float volume) { impl_->set_master_volume(volume); }
float audio_mixer::get_master_volume() { return impl_->get_master_volume(); }
-audio_buffer audio_mixer::operator()(const video_format_desc& format_desc){ return impl_->mix(format_desc); }
+audio_buffer audio_mixer::operator()(const video_format_desc& format_desc, const audio_channel_layout& channel_layout){ return impl_->mix(format_desc, channel_layout); }
monitor::subject& audio_mixer::monitor_output(){ return impl_->monitor_subject_; }
}}