28f670f09f
The algorithm is ported over from Chromium's, which produces very good results for speech, while being not objectionable for music, especially in the background. Other algorithms were tested. Phase vocoders: - Bungee - Signalsmith Stretch - pvdoneright All three of these exhibited the usual phase shifting artifacts, causing speech to sound slightly shifted into the high end. Speech is the main thing we want to optimize for, so these aren't ideal. Sonic (TD-PSOLA) performs better than WSOLA for speech, especially at rates higher than 2x, but makes background sounds/music garbled and unpleasant. It is still worth considering for speech clarity, and could be added as an optional feature.
85 lines
2.4 KiB
C++
85 lines
2.4 KiB
C++
/*
|
|
* Copyright 2012 The Chromium Authors
|
|
* Copyright (c) 2026-present, the Ladybird developers.
|
|
*
|
|
* SPDX-License-Identifier: BSD-3-Clause
|
|
*/
|
|
// Source: chromium/media/filters/audio_renderer_algorithm.{h,cc}.
|
|
|
|
#pragma once
|
|
|
|
#include <AK/Vector.h>
|
|
#include <LibMedia/Audio/AudioBuffer.h>
|
|
#include <LibMedia/Audio/SampleSpecification.h>
|
|
#include <LibMedia/AudioBlock.h>
|
|
|
|
namespace Audio {
|
|
|
|
class WSOLAAlgorithm {
|
|
public:
|
|
explicit WSOLAAlgorithm(SampleSpecification);
|
|
~WSOLAAlgorithm();
|
|
|
|
WSOLAAlgorithm(WSOLAAlgorithm const&) = delete;
|
|
WSOLAAlgorithm& operator=(WSOLAAlgorithm const&) = delete;
|
|
|
|
size_t fill_buffer(Media::AudioBlock& destination, size_t destination_offset, size_t requested_frames, float playback_rate);
|
|
|
|
void enqueue_buffer(Media::AudioBlock const&);
|
|
|
|
void flush_buffers();
|
|
|
|
void mark_end_of_stream(float playback_rate);
|
|
|
|
size_t buffered_frames() const { return m_audio_buffer.frame_count(); }
|
|
|
|
SampleSpecification const& sample_specification() const { return m_sample_specification; }
|
|
u32 sample_rate() const { return m_sample_specification.sample_rate(); }
|
|
u8 channel_count() const { return m_sample_specification.channel_count(); }
|
|
size_t ola_window_size() const { return m_ola_window_size; }
|
|
size_t ola_hop_size() const { return m_ola_hop_size; }
|
|
|
|
private:
|
|
bool run_one_wsola_iteration(float playback_rate);
|
|
|
|
size_t write_completed_frames_to(size_t requested_frames, size_t destination_offset, Media::AudioBlock&);
|
|
|
|
void peek_audio_with_zero_prepend(i64 read_offset_frames, Media::AudioBlock& destination);
|
|
|
|
void get_optimal_block();
|
|
|
|
bool can_perform_wsola() const;
|
|
|
|
bool target_is_within_search_region() const;
|
|
|
|
void remove_old_input_frames();
|
|
|
|
Media::AudioBlock create_block(size_t frame_count) const;
|
|
|
|
SampleSpecification const m_sample_specification;
|
|
|
|
AudioBuffer m_audio_buffer;
|
|
|
|
size_t const m_ola_window_size;
|
|
size_t const m_ola_hop_size;
|
|
size_t const m_num_candidate_blocks;
|
|
size_t const m_search_block_center_offset;
|
|
|
|
float m_input_position_remainder { 0.0f };
|
|
i64 m_search_block_index { 0 };
|
|
i64 m_search_block_center_index { 0 };
|
|
i64 m_target_block_index { 0 };
|
|
|
|
size_t m_num_complete_frames { 0 };
|
|
|
|
Media::AudioBlock m_wsola_output;
|
|
|
|
Vector<float> m_ola_window;
|
|
Vector<float> m_transition_window;
|
|
|
|
Media::AudioBlock m_optimal_block;
|
|
Media::AudioBlock m_search_block;
|
|
Media::AudioBlock m_target_block;
|
|
};
|
|
|
|
}
|