AudioRecorder.cpp

Go to the documentation of this file.

 
// Copyright (c) 2008-2026 OpenShot Studios, LLC
//
// SPDX-License-Identifier: LGPL-3.0-or-later
 
#include "AudioRecorder.h"
 
#include <algorithm>
#include <cmath>
#include <utility>
 
#include "Exceptions.h"
#include "FFmpegWriter.h"
#include "Frame.h"
 
using namespace openshot;
 
int AudioRecorderInputMapper::DeviceChannels(int requested_channels)
{
    return std::max(requested_channels, 2);
}
 
void AudioRecorderInputMapper::FillBlock(
    AudioRecorderBlock& block,
    int channels,
    const float* const* input_channel_data,
    int num_input_channels,
    int num_samples)
{
    block.channels.resize(channels);
    for (int channel = 0; channel < channels; ++channel) {
        block.channels[channel].assign(num_samples, 0.0f);
        const int source = channel < num_input_channels ? channel : 0;
        if (source < num_input_channels && input_channel_data[source]) {
            std::copy(input_channel_data[source], input_channel_data[source] + num_samples, block.channels[channel].begin());
        }
    }
}
 
AudioLevelData AudioRecorderLevelMeter::ProcessBlock(const AudioRecorderBlock& block) const
{
    AudioLevelData result;
    result.timestamp = block.sample_rate > 0
        ? static_cast<double>(block.first_sample) / static_cast<double>(block.sample_rate)
        : 0.0;
 
    const int channels = static_cast<int>(block.channels.size());
    result.peak.assign(channels, 0.0f);
    result.rms.assign(channels, 0.0f);
 
    for (int channel = 0; channel < channels; ++channel) {
        const auto& samples = block.channels[channel];
        double squared_sum = 0.0;
        float peak = 0.0f;
 
        for (float sample : samples) {
            const float abs_sample = std::abs(sample);
            peak = std::max(peak, abs_sample);
            squared_sum += static_cast<double>(sample) * static_cast<double>(sample);
            if (abs_sample >= 1.0f) {
                result.clipped = true;
            }
        }
 
        result.peak[channel] = peak;
        if (!samples.empty()) {
            result.rms[channel] = static_cast<float>(std::sqrt(squared_sum / static_cast<double>(samples.size())));
        }
    }
 
    return result;
}
 
AudioRecorderWaveformAccumulator::AudioRecorderWaveformAccumulator(int new_sample_rate, int new_samples_per_second)
    : sample_rate(new_sample_rate)
    , samples_per_second(new_samples_per_second)
    , sample_divisor(1)
    , pending_samples(0)
    , pending_max(0.0f)
    , pending_squared_sum(0.0)
    , emitted_visual_samples(0)
{
    if (sample_rate <= 0 || samples_per_second <= 0) {
        throw InvalidOptions("Audio waveform settings require a valid sample rate and samples-per-second value.");
    }
 
    sample_divisor = std::max(1, sample_rate / samples_per_second);
}
 
std::vector<AudioWaveformChunk> AudioRecorderWaveformAccumulator::ProcessBlock(const AudioRecorderBlock& block)
{
    std::vector<AudioWaveformChunk> chunks;
    if (block.channels.empty() || block.Samples() <= 0) {
        return chunks;
    }
 
    AudioWaveformChunk chunk;
    chunk.samples_per_second = samples_per_second;
    chunk.start_time = static_cast<double>(emitted_visual_samples) / static_cast<double>(samples_per_second);
 
    const int channels = static_cast<int>(block.channels.size());
    const int samples = block.Samples();
    for (int sample_index = 0; sample_index < samples; ++sample_index) {
        for (int channel = 0; channel < channels; ++channel) {
            if (sample_index >= static_cast<int>(block.channels[channel].size())) {
                continue;
            }
            const float sample = block.channels[channel][sample_index];
            pending_max = std::max(pending_max, std::abs(sample));
            pending_squared_sum += static_cast<double>(sample) * static_cast<double>(sample);
        }
 
        pending_samples++;
        if (pending_samples >= sample_divisor) {
            const double denominator = static_cast<double>(pending_samples * channels);
            const float rms = denominator > 0.0
                ? static_cast<float>(std::sqrt(pending_squared_sum / denominator))
                : 0.0f;
 
            max_samples.push_back(pending_max);
            rms_samples.push_back(rms);
            chunk.max_samples.push_back(pending_max);
            chunk.rms_samples.push_back(rms);
            emitted_visual_samples++;
 
            pending_samples = 0;
            pending_max = 0.0f;
            pending_squared_sum = 0.0;
        }
    }
 
    if (!chunk.max_samples.empty()) {
        chunk.duration = static_cast<double>(chunk.max_samples.size()) / static_cast<double>(samples_per_second);
        chunks.push_back(std::move(chunk));
    }
 
    return chunks;
}
 
AudioWaveformData AudioRecorderWaveformAccumulator::Snapshot() const
{
    AudioWaveformData result;
    result.max_samples = max_samples;
    result.rms_samples = rms_samples;
    return result;
}
 
void AudioRecorderWaveformAccumulator::Reset()
{
    pending_samples = 0;
    pending_max = 0.0f;
    pending_squared_sum = 0.0;
    emitted_visual_samples = 0;
    max_samples.clear();
    rms_samples.clear();
}
 
std::shared_ptr<Frame> AudioRecordingFrameFactory::CreateFrame(
    const AudioRecorderBlock& block,
    ChannelLayout channel_layout,
    int64_t frame_number)
{
    const int channels = static_cast<int>(block.channels.size());
    const int samples = block.Samples();
    auto frame = std::make_shared<Frame>(frame_number, samples, channels);
    frame->SampleRate(block.sample_rate);
    frame->ChannelsLayout(channel_layout);
 
    for (int channel = 0; channel < channels; ++channel) {
        frame->AddAudio(true, channel, 0, block.channels[channel].data(), samples, 1.0f);
    }
 
    return frame;
}
 
AudioRecorder::AudioRecorder(const AudioRecorderSettings& new_settings)
    : settings(new_settings)
    , writer(nullptr)
    , waveform_accumulator(nullptr)
    , is_open(false)
    , is_recording(false)
    , is_monitoring(false)
    , writer_should_stop(false)
    , samples_recorded(0)
    , dropped_blocks(0)
    , next_frame_number(1)
{
    ValidateSettings();
}
 
AudioRecorder::~AudioRecorder()
{
    Close();
}
 
void AudioRecorder::ValidateSettings() const
{
    if (settings.path.empty()) {
        throw InvalidOptions("Audio recorder requires an output path.");
    }
    if (settings.codec.empty()) {
        throw InvalidOptions("Audio recorder requires an audio codec.");
    }
    if (settings.sample_rate < 8000) {
        throw InvalidSampleRate("Audio recorder requires a sample rate of at least 8000 Hz.", settings.path);
    }
    if (settings.channels <= 0) {
        throw InvalidChannels("Audio recorder requires at least one input channel.", settings.path);
    }
    if (settings.buffer_size <= 0) {
        throw InvalidOptions("Audio recorder requires a positive audio buffer size.", settings.path);
    }
    if (settings.waveform_samples_per_second <= 0) {
        throw InvalidOptions("Audio recorder requires a positive waveform sample rate.", settings.path);
    }
    if (settings.max_queue_seconds <= 0) {
        throw InvalidOptions("Audio recorder requires a positive maximum queue duration.", settings.path);
    }
}
 
void AudioRecorder::Open()
{
    if (is_open) {
        return;
    }
 
    if (!settings.device_type.empty()) {
        // Device types are created lazily. Selecting a type before the scan is
        // a no-op, which can send a DirectSound input name to WASAPI on Windows.
        device_manager.getAvailableDeviceTypes();
        device_manager.setCurrentAudioDeviceType(settings.device_type, true);
    }
 
    // Open stereo devices as stereo, and pick the requested channels ourselves.
    // If a driver refuses that, fall back to the requested channel count.
    const int device_channels = AudioRecorderInputMapper::DeviceChannels(settings.channels);
    juce::String error = OpenDevice(device_channels);
    if (error.isNotEmpty() && device_channels != settings.channels) {
        device_manager.closeAudioDevice();
        error = OpenDevice(settings.channels);
    }
    if (error.isNotEmpty()) {
        throw InvalidOptions(error.toStdString(), settings.path);
    }
 
    if (auto* device = device_manager.getCurrentAudioDevice()) {
        const double actual_rate = device->getCurrentSampleRate();
        if (actual_rate > 0.0) {
            settings.sample_rate = static_cast<int>(std::llround(actual_rate));
        }
    }
 
    waveform_accumulator = std::make_unique<AudioRecorderWaveformAccumulator>(
        settings.sample_rate,
        settings.waveform_samples_per_second);
 
    is_open = true;
}
 
juce::String AudioRecorder::OpenDevice(int device_channels)
{
    juce::AudioDeviceManager::AudioDeviceSetup setup;
    setup.inputChannels.clear();
    setup.inputChannels.setRange(0, device_channels, true);
    setup.outputChannels.clear();
    setup.sampleRate = settings.sample_rate;
    setup.bufferSize = settings.buffer_size;
    setup.inputDeviceName = settings.device_name;
 
    return device_manager.initialise(
        device_channels,
        0,
        nullptr,
        true,
        settings.device_name,
        &setup);
}
 
void AudioRecorder::OpenWriter()
{
    if (writer) {
        return;
    }
 
    writer = std::make_unique<FFmpegWriter>(settings.path);
    writer->SetAudioOptions(true, settings.codec, settings.sample_rate, settings.channels, settings.channel_layout, settings.bit_rate);
    writer->Open();
}
 
void AudioRecorder::Start()
{
    if (is_recording) {
        return;
    }
 
    PrepareRecording();
 
    writer_should_stop = false;
    is_recording = true;
    device_manager.addAudioCallback(this);
    writer_thread = std::thread(&AudioRecorder::WriterLoop, this);
}
 
void AudioRecorder::PrepareRecording()
{
    if (!is_open) {
        Open();
    }
    StopMonitoring();
    OpenWriter();
    if (waveform_accumulator) {
        waveform_accumulator->Reset();
    }
    samples_recorded = 0;
    dropped_blocks = 0;
    next_frame_number = 1;
}
 
void AudioRecorder::Stop()
{
    if (!is_recording && !writer_thread.joinable()) {
        if (writer) {
            writer->Close();
            writer.reset();
        }
        return;
    }
 
    is_recording = false;
    device_manager.removeAudioCallback(this);
    writer_should_stop = true;
    queue_condition.notify_all();
 
    if (writer_thread.joinable()) {
        writer_thread.join();
    }
 
    if (writer) {
        writer->Close();
        writer.reset();
    }
}
 
void AudioRecorder::StartMonitoring()
{
    if (is_recording || is_monitoring) {
        return;
    }
 
    if (!is_open) {
        Open();
    }
 
    is_monitoring = true;
    device_manager.addAudioCallback(this);
}
 
void AudioRecorder::StopMonitoring()
{
    if (!is_monitoring) {
        return;
    }
 
    is_monitoring = false;
    device_manager.removeAudioCallback(this);
}
 
void AudioRecorder::Close()
{
    StopMonitoring();
    Stop();
 
    if (is_open) {
        device_manager.closeAudioDevice();
        if (writer) {
            writer->Close();
            writer.reset();
        }
        is_open = false;
    }
}
 
bool AudioRecorder::IsOpen() const
{
    return is_open;
}
 
bool AudioRecorder::IsRecording() const
{
    return is_recording;
}
 
bool AudioRecorder::IsMonitoring() const
{
    return is_monitoring;
}
 
AudioRecorderStats AudioRecorder::GetStats() const
{
    AudioRecorderStats stats;
    stats.is_open = is_open;
    stats.is_recording = is_recording;
    stats.sample_rate = settings.sample_rate;
    stats.channels = settings.channels;
    stats.samples_recorded = samples_recorded;
    stats.dropped_blocks = dropped_blocks;
    stats.duration = settings.sample_rate > 0
        ? static_cast<double>(stats.samples_recorded) / static_cast<double>(settings.sample_rate)
        : 0.0;
 
    std::lock_guard<std::mutex> lock(queue_mutex);
    stats.queued_blocks = static_cast<int64_t>(queue.size());
    return stats;
}
 
AudioWaveformData AudioRecorder::GetWaveformSnapshot() const
{
    std::lock_guard<std::mutex> lock(state_mutex);
    return waveform_accumulator ? waveform_accumulator->Snapshot() : AudioWaveformData();
}
 
AudioLevelData AudioRecorder::GetLevelSnapshot() const
{
    std::lock_guard<std::mutex> lock(state_mutex);
    return last_level;
}
 
void AudioRecorder::SetLevelCallback(std::function<void(const AudioLevelData&)> callback)
{
    std::lock_guard<std::mutex> lock(state_mutex);
    level_callback = std::move(callback);
}
 
void AudioRecorder::SetWaveformCallback(std::function<void(const AudioWaveformChunk&)> callback)
{
    std::lock_guard<std::mutex> lock(state_mutex);
    waveform_callback = std::move(callback);
}
 
void AudioRecorder::audioDeviceIOCallbackWithContext(
    const float* const* inputChannelData,
    int numInputChannels,
    float* const* outputChannelData,
    int numOutputChannels,
    int numSamples,
    const juce::AudioIODeviceCallbackContext&)
{
    for (int channel = 0; channel < numOutputChannels; ++channel) {
        if (outputChannelData[channel]) {
            std::fill(outputChannelData[channel], outputChannelData[channel] + numSamples, 0.0f);
        }
    }
 
    if ((!is_recording && !is_monitoring) || numSamples <= 0) {
        return;
    }
 
    AudioRecorderBlock block;
    block.sample_rate = settings.sample_rate;
    block.first_sample = samples_recorded.load();
    AudioRecorderInputMapper::FillBlock(block, settings.channels, inputChannelData, numInputChannels, numSamples);
 
    AudioLevelData level = level_meter.ProcessBlock(block);
    std::function<void(const AudioLevelData&)> level_cb;
    {
        std::lock_guard<std::mutex> lock(state_mutex);
        last_level = level;
        level_cb = level_callback;
    }
    if (level_cb) {
        level_cb(level);
    }
 
    if (!is_recording) {
        samples_recorded += numSamples;
        return;
    }
 
    const int64_t max_queue_blocks = static_cast<int64_t>(
        std::max(1, (settings.max_queue_seconds * settings.sample_rate) / std::max(1, numSamples)));
    {
        std::lock_guard<std::mutex> lock(queue_mutex);
        if (static_cast<int64_t>(queue.size()) >= max_queue_blocks) {
            dropped_blocks++;
        } else {
            queue.push_back(std::move(block));
            queue_condition.notify_one();
        }
    }
 
    samples_recorded += numSamples;
}
 
void AudioRecorder::audioDeviceAboutToStart(juce::AudioIODevice* device)
{
    (void) device;
}
 
void AudioRecorder::audioDeviceStopped()
{
}
 
bool AudioRecorder::PopBlock(AudioRecorderBlock& block)
{
    std::unique_lock<std::mutex> lock(queue_mutex);
    queue_condition.wait(lock, [this]() {
        return writer_should_stop || !queue.empty();
    });
 
    if (queue.empty()) {
        return false;
    }
 
    block = std::move(queue.front());
    queue.pop_front();
    return true;
}
 
void AudioRecorder::WriterLoop()
{
    int64_t expected_next_sample = -1;
    while (true) {
        AudioRecorderBlock block;
        if (!PopBlock(block)) {
            if (writer_should_stop) {
                break;
            }
            continue;
        }
 
        std::vector<AudioWaveformChunk> waveform_chunks;
        std::function<void(const AudioWaveformChunk&)> waveform_cb;
        {
            std::lock_guard<std::mutex> lock(state_mutex);
            if (waveform_accumulator) {
                waveform_chunks = waveform_accumulator->ProcessBlock(block);
            }
            waveform_cb = waveform_callback;
        }
 
        if (waveform_cb) {
            for (const auto& chunk : waveform_chunks) {
                waveform_cb(chunk);
            }
        }
 
        if (writer) {
            while (expected_next_sample >= 0 && block.first_sample > expected_next_sample) {
                const int64_t missing_samples = block.first_sample - expected_next_sample;
                const int silence_samples = static_cast<int>(std::min<int64_t>(
                    missing_samples,
                    std::max(1, settings.buffer_size)));
                AudioRecorderBlock silence_block;
                silence_block.sample_rate = block.sample_rate;
                silence_block.first_sample = expected_next_sample;
                silence_block.channels.assign(
                    settings.channels,
                    std::vector<float>(silence_samples, 0.0f));
                writer->WriteFrame(AudioRecordingFrameFactory::CreateFrame(
                    silence_block,
                    settings.channel_layout,
                    next_frame_number++));
                expected_next_sample += silence_samples;
            }
 
            writer->WriteFrame(AudioRecordingFrameFactory::CreateFrame(
                block,
                settings.channel_layout,
                next_frame_number++));
        }
        expected_next_sample = block.first_sample + block.Samples();
    }
}