/*************************************************************************/ /* audio_effect_chorus.cpp */ /*************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ /* https://godotengine.org */ /*************************************************************************/ /* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ /* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ /* */ /* Permission is hereby granted, free of charge, to any person obtaining */ /* a copy of this software and associated documentation files (the */ /* "Software"), to deal in the Software without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of the Software, and to */ /* permit persons to whom the Software is furnished to do so, subject to */ /* the following conditions: */ /* */ /* The above copyright notice and this permission notice shall be */ /* included in all copies or substantial portions of the Software. */ /* */ /* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ /* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ /* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ /* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ /* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ /* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ #include "audio_effect_chorus.h" #include "core/math/math_funcs.h" #include "servers/audio_server.h" void AudioEffectChorusInstance::process(const AudioFrame *p_src_frames, AudioFrame *p_dst_frames, int p_frame_count) { int todo = p_frame_count; while (todo) { int to_mix = MIN(todo, 256); //can't mix too much _process_chunk(p_src_frames, p_dst_frames, to_mix); p_src_frames += to_mix; p_dst_frames += to_mix; todo -= to_mix; } } void AudioEffectChorusInstance::_process_chunk(const AudioFrame *p_src_frames, AudioFrame *p_dst_frames, int p_frame_count) { //fill ringbuffer for (int i = 0; i < p_frame_count; i++) { audio_buffer.write[(buffer_pos + i) & buffer_mask] = p_src_frames[i]; p_dst_frames[i] = p_src_frames[i] * base->dry; } float mix_rate = AudioServer::get_singleton()->get_mix_rate(); /* process voices */ for (int vc = 0; vc < base->voice_count; vc++) { AudioEffectChorus::Voice &v = base->voice[vc]; double time_to_mix = (float)p_frame_count / mix_rate; double cycles_to_mix = time_to_mix * v.rate; unsigned int local_rb_pos = buffer_pos; AudioFrame *dst_buff = p_dst_frames; AudioFrame *rb_buff = audio_buffer.ptrw(); double delay_msec = v.delay; unsigned int delay_frames = Math::fast_ftoi((delay_msec / 1000.0) * mix_rate); float max_depth_frames = (v.depth / 1000.0) * mix_rate; uint64_t local_cycles = cycles[vc]; uint64_t increment = llrint(cycles_to_mix / (double)p_frame_count * (double)(1 << AudioEffectChorus::CYCLES_FRAC)); //check the LFO doesn't read ahead of the write pos if ((((unsigned int)max_depth_frames) + 10) > delay_frames) { //10 as some threshold to avoid precision stuff delay_frames += (int)max_depth_frames - delay_frames; delay_frames += 10; //threshold to avoid precision stuff } //low pass filter if (v.cutoff == 0) { continue; } float auxlp = expf(-2.0 * Math_PI * v.cutoff / mix_rate); float c1 = 1.0 - auxlp; float c2 = auxlp; AudioFrame h = filter_h[vc]; if (v.cutoff >= AudioEffectChorus::MS_CUTOFF_MAX) { c1 = 1.0; c2 = 0.0; } //vol modifier AudioFrame vol_modifier = AudioFrame(base->wet, base->wet) * Math::db2linear(v.level); vol_modifier.l *= CLAMP(1.0 - v.pan, 0, 1); vol_modifier.r *= CLAMP(1.0 + v.pan, 0, 1); for (int i = 0; i < p_frame_count; i++) { /** COMPUTE WAVEFORM **/ float phase = (float)(local_cycles & AudioEffectChorus::CYCLES_MASK) / (float)(1 << AudioEffectChorus::CYCLES_FRAC); float wave_delay = sinf(phase * 2.0 * Math_PI) * max_depth_frames; int wave_delay_frames = lrint(floor(wave_delay)); float wave_delay_frac = wave_delay - (float)wave_delay_frames; /** COMPUTE RINGBUFFER POS**/ unsigned int rb_source = local_rb_pos; rb_source -= delay_frames; rb_source -= wave_delay_frames; /** READ FROM RINGBUFFER, LINEARLY INTERPOLATE */ AudioFrame val = rb_buff[rb_source & buffer_mask]; AudioFrame val_next = rb_buff[(rb_source - 1) & buffer_mask]; val += (val_next - val) * wave_delay_frac; val = val * c1 + h * c2; h = val; /** MIX VALUE TO OUTPUT **/ dst_buff[i] += val * vol_modifier; local_cycles += increment; local_rb_pos++; } filter_h[vc] = h; cycles[vc] += Math::fast_ftoi(cycles_to_mix * (double)(1 << AudioEffectChorus::CYCLES_FRAC)); } buffer_pos += p_frame_count; } Ref AudioEffectChorus::instance() { Ref ins; ins.instance(); ins->base = Ref(this); for (int i = 0; i < 4; i++) { ins->filter_h[i] = AudioFrame(0, 0); ins->cycles[i] = 0; } float ring_buffer_max_size = AudioEffectChorus::MAX_DELAY_MS + AudioEffectChorus::MAX_DEPTH_MS + AudioEffectChorus::MAX_WIDTH_MS; ring_buffer_max_size *= 2; //just to avoid complications ring_buffer_max_size /= 1000.0; //convert to seconds ring_buffer_max_size *= AudioServer::get_singleton()->get_mix_rate(); int ringbuff_size = ring_buffer_max_size; int bits = 0; while (ringbuff_size > 0) { bits++; ringbuff_size /= 2; } ringbuff_size = 1 << bits; ins->buffer_mask = ringbuff_size - 1; ins->buffer_pos = 0; ins->audio_buffer.resize(ringbuff_size); for (int i = 0; i < ringbuff_size; i++) { ins->audio_buffer.write[i] = AudioFrame(0, 0); } return ins; } void AudioEffectChorus::set_voice_count(int p_voices) { ERR_FAIL_COND(p_voices < 1 || p_voices > MAX_VOICES); voice_count = p_voices; } int AudioEffectChorus::get_voice_count() const { return voice_count; } void AudioEffectChorus::set_voice_delay_ms(int p_voice, float p_delay_ms) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].delay = p_delay_ms; } float AudioEffectChorus::get_voice_delay_ms(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].delay; } void AudioEffectChorus::set_voice_rate_hz(int p_voice, float p_rate_hz) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].rate = p_rate_hz; } float AudioEffectChorus::get_voice_rate_hz(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].rate; } void AudioEffectChorus::set_voice_depth_ms(int p_voice, float p_depth_ms) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].depth = p_depth_ms; } float AudioEffectChorus::get_voice_depth_ms(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].depth; } void AudioEffectChorus::set_voice_level_db(int p_voice, float p_level_db) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].level = p_level_db; } float AudioEffectChorus::get_voice_level_db(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].level; } void AudioEffectChorus::set_voice_cutoff_hz(int p_voice, float p_cutoff_hz) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].cutoff = p_cutoff_hz; } float AudioEffectChorus::get_voice_cutoff_hz(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].cutoff; } void AudioEffectChorus::set_voice_pan(int p_voice, float p_pan) { ERR_FAIL_INDEX(p_voice, MAX_VOICES); voice[p_voice].pan = p_pan; } float AudioEffectChorus::get_voice_pan(int p_voice) const { ERR_FAIL_INDEX_V(p_voice, MAX_VOICES, 0); return voice[p_voice].pan; } void AudioEffectChorus::set_wet(float amount) { wet = amount; } float AudioEffectChorus::get_wet() const { return wet; } void AudioEffectChorus::set_dry(float amount) { dry = amount; } float AudioEffectChorus::get_dry() const { return dry; } void AudioEffectChorus::_validate_property(PropertyInfo &property) const { if (property.name.begins_with("voice/")) { int voice_idx = property.name.get_slice("/", 1).to_int(); if (voice_idx > voice_count) { property.usage = 0; } } } void AudioEffectChorus::_bind_methods() { ClassDB::bind_method(D_METHOD("set_voice_count", "voices"), &AudioEffectChorus::set_voice_count); ClassDB::bind_method(D_METHOD("get_voice_count"), &AudioEffectChorus::get_voice_count); ClassDB::bind_method(D_METHOD("set_voice_delay_ms", "voice_idx", "delay_ms"), &AudioEffectChorus::set_voice_delay_ms); ClassDB::bind_method(D_METHOD("get_voice_delay_ms", "voice_idx"), &AudioEffectChorus::get_voice_delay_ms); ClassDB::bind_method(D_METHOD("set_voice_rate_hz", "voice_idx", "rate_hz"), &AudioEffectChorus::set_voice_rate_hz); ClassDB::bind_method(D_METHOD("get_voice_rate_hz", "voice_idx"), &AudioEffectChorus::get_voice_rate_hz); ClassDB::bind_method(D_METHOD("set_voice_depth_ms", "voice_idx", "depth_ms"), &AudioEffectChorus::set_voice_depth_ms); ClassDB::bind_method(D_METHOD("get_voice_depth_ms", "voice_idx"), &AudioEffectChorus::get_voice_depth_ms); ClassDB::bind_method(D_METHOD("set_voice_level_db", "voice_idx", "level_db"), &AudioEffectChorus::set_voice_level_db); ClassDB::bind_method(D_METHOD("get_voice_level_db", "voice_idx"), &AudioEffectChorus::get_voice_level_db); ClassDB::bind_method(D_METHOD("set_voice_cutoff_hz", "voice_idx", "cutoff_hz"), &AudioEffectChorus::set_voice_cutoff_hz); ClassDB::bind_method(D_METHOD("get_voice_cutoff_hz", "voice_idx"), &AudioEffectChorus::get_voice_cutoff_hz); ClassDB::bind_method(D_METHOD("set_voice_pan", "voice_idx", "pan"), &AudioEffectChorus::set_voice_pan); ClassDB::bind_method(D_METHOD("get_voice_pan", "voice_idx"), &AudioEffectChorus::get_voice_pan); ClassDB::bind_method(D_METHOD("set_wet", "amount"), &AudioEffectChorus::set_wet); ClassDB::bind_method(D_METHOD("get_wet"), &AudioEffectChorus::get_wet); ClassDB::bind_method(D_METHOD("set_dry", "amount"), &AudioEffectChorus::set_dry); ClassDB::bind_method(D_METHOD("get_dry"), &AudioEffectChorus::get_dry); ADD_PROPERTY(PropertyInfo(Variant::INT, "voice_count", PROPERTY_HINT_RANGE, "1,4,1", PROPERTY_USAGE_DEFAULT | PROPERTY_USAGE_UPDATE_ALL_IF_MODIFIED), "set_voice_count", "get_voice_count"); ADD_PROPERTY(PropertyInfo(Variant::REAL, "dry", PROPERTY_HINT_RANGE, "0,1,0.01"), "set_dry", "get_dry"); ADD_PROPERTY(PropertyInfo(Variant::REAL, "wet", PROPERTY_HINT_RANGE, "0,1,0.01"), "set_wet", "get_wet"); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/delay_ms", PROPERTY_HINT_RANGE, "0,50,0.01"), "set_voice_delay_ms", "get_voice_delay_ms", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/rate_hz", PROPERTY_HINT_RANGE, "0.1,20,0.1"), "set_voice_rate_hz", "get_voice_rate_hz", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/depth_ms", PROPERTY_HINT_RANGE, "0,20,0.01"), "set_voice_depth_ms", "get_voice_depth_ms", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/level_db", PROPERTY_HINT_RANGE, "-60,24,0.1"), "set_voice_level_db", "get_voice_level_db", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/cutoff_hz", PROPERTY_HINT_RANGE, "1,20500,1"), "set_voice_cutoff_hz", "get_voice_cutoff_hz", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/1/pan", PROPERTY_HINT_RANGE, "-1,1,0.01"), "set_voice_pan", "get_voice_pan", 0); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/delay_ms", PROPERTY_HINT_RANGE, "0,50,0.01"), "set_voice_delay_ms", "get_voice_delay_ms", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/rate_hz", PROPERTY_HINT_RANGE, "0.1,20,0.1"), "set_voice_rate_hz", "get_voice_rate_hz", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/depth_ms", PROPERTY_HINT_RANGE, "0,20,0.01"), "set_voice_depth_ms", "get_voice_depth_ms", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/level_db", PROPERTY_HINT_RANGE, "-60,24,0.1"), "set_voice_level_db", "get_voice_level_db", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/cutoff_hz", PROPERTY_HINT_RANGE, "1,20500,1"), "set_voice_cutoff_hz", "get_voice_cutoff_hz", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/2/pan", PROPERTY_HINT_RANGE, "-1,1,0.01"), "set_voice_pan", "get_voice_pan", 1); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/delay_ms", PROPERTY_HINT_RANGE, "0,50,0.01"), "set_voice_delay_ms", "get_voice_delay_ms", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/rate_hz", PROPERTY_HINT_RANGE, "0.1,20,0.1"), "set_voice_rate_hz", "get_voice_rate_hz", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/depth_ms", PROPERTY_HINT_RANGE, "0,20,0.01"), "set_voice_depth_ms", "get_voice_depth_ms", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/level_db", PROPERTY_HINT_RANGE, "-60,24,0.1"), "set_voice_level_db", "get_voice_level_db", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/cutoff_hz", PROPERTY_HINT_RANGE, "1,20500,1"), "set_voice_cutoff_hz", "get_voice_cutoff_hz", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/3/pan", PROPERTY_HINT_RANGE, "-1,1,0.01"), "set_voice_pan", "get_voice_pan", 2); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/delay_ms", PROPERTY_HINT_RANGE, "0,50,0.01"), "set_voice_delay_ms", "get_voice_delay_ms", 3); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/rate_hz", PROPERTY_HINT_RANGE, "0.1,20,0.1"), "set_voice_rate_hz", "get_voice_rate_hz", 3); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/depth_ms", PROPERTY_HINT_RANGE, "0,20,0.01"), "set_voice_depth_ms", "get_voice_depth_ms", 3); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/level_db", PROPERTY_HINT_RANGE, "-60,24,0.1"), "set_voice_level_db", "get_voice_level_db", 3); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/cutoff_hz", PROPERTY_HINT_RANGE, "1,20500,1"), "set_voice_cutoff_hz", "get_voice_cutoff_hz", 3); ADD_PROPERTYI(PropertyInfo(Variant::REAL, "voice/4/pan", PROPERTY_HINT_RANGE, "-1,1,0.01"), "set_voice_pan", "get_voice_pan", 3); } AudioEffectChorus::AudioEffectChorus() { voice_count = 2; voice[0].delay = 15; voice[1].delay = 20; voice[0].rate = 0.8; voice[1].rate = 1.2; voice[0].depth = 2; voice[1].depth = 3; voice[0].cutoff = 8000; voice[1].cutoff = 8000; voice[0].pan = -0.5; voice[1].pan = 0.5; wet = 0.5; dry = 1.0; }