Function mix_audio_to_reference_shape

src/framework/audio/mixing.cpp:190–231

}  // namespace

WavData mix_audio_to_reference_shape(
    const std::vector<AudioMixInput> & inputs,
    bool normalize_peak) {
    if (inputs.empty()) {
        throw std::runtime_error("audio mix requires at least one input");
    }
    const WavData & reference = inputs.front().audio;
    const int64_t frames = frame_count(reference);
    if (reference.sample_rate <= 0 || reference.channels <= 0 || frames <= 0) {
        throw std::runtime_error("audio mix reference input is invalid");
    }
    WavData out;
    out.sample_rate = reference.sample_rate;
    out.channels = reference.channels;
    out.samples.assign(static_cast<size_t>(frames * reference.channels), 0.0F);
    for (const auto & input : inputs) {
        auto adapted = adapt_audio(input.audio, out.sample_rate, out.channels, frames);
        if (input.activity_reference.has_value()) {
            auto reference_activity = adapt_audio(*input.activity_reference, out.sample_rate, out.channels, frames);
            apply_activity_gate_in_place(
                adapted,
                reference_activity,
                out.channels,
                out.sample_rate,
                input.activity_threshold_dbfs,
                input.activity_window_seconds,
                input.activity_margin_seconds,
                input.activity_fade_seconds);
        }
        for (size_t i = 0; i < out.samples.size(); ++i) {
            out.samples[i] += adapted[i] * input.gain;
        }
    }
    if (normalize_peak) {
        normalize_peak_to_unit_range_and_clamp_in_place(out.samples);
    } else {
        for (float & sample : out.samples) {
            sample = std::clamp(sample, -1.0F, 1.0F);
        }
    }
    return out;
}

WavData read_and_mix_wavs_to_reference_shape(
    const std::vector<std::pair<std::filesystem::path, float>> & inputs,