Merge pull request #91014 from DeeJayLSP/qoa-wav-playback

Add QOA (Quite OK Audio) as a WAV compression mode

Author: akien-mga

COPYRIGHT.txt

@@ -411,6 +411,11 @@ Comment: PolyPartition / Triangulator
 Copyright: 2011-2021, Ivan Fratric and contributors
 License: Expat
 
+Files: ./thirdparty/misc/qoa.h
+Comment: Quite OK Audio Format
+Copyright: 2023, Dominic Szablewski
+License: Expat
+
 Files: ./thirdparty/misc/r128.c
  ./thirdparty/misc/r128.h
 Comment: r128 library

doc/classes/AudioStreamWAV.xml

@@ -15,7 +15,7 @@
 			<return type="int" enum="Error" />
 			<param index="0" name="path" type="String" />
 			<description>
-				Saves the AudioStreamWAV as a WAV file to [param path]. Samples with IMA ADPCM format can't be saved.
+				Saves the AudioStreamWAV as a WAV file to [param path]. Samples with IMA ADPCM or QOA formats can't be saved.
 				[b]Note:[/b] A [code].wav[/code] extension is automatically appended to [param path] if it is missing.
 			</description>
 		</method>
@@ -56,6 +56,9 @@
 		<constant name="FORMAT_IMA_ADPCM" value="2" enum="Format">
 			Audio is compressed using IMA ADPCM.
 		</constant>
+		<constant name="FORMAT_QOA" value="3" enum="Format">
+			Audio is compressed as QOA ([url=https://qoaformat.org/]Quite OK Audio[/url]).
+		</constant>
 		<constant name="LOOP_DISABLED" value="0" enum="LoopMode">
 			Audio does not loop.
 		</constant>

doc/classes/ResourceImporterWAV.xml

@@ -14,6 +14,7 @@
 			The compression mode to use on import.
 			[b]Disabled:[/b] Imports audio data without any compression. This results in the highest possible quality.
 			[b]RAM (Ima-ADPCM):[/b] Performs fast lossy compression on import. Low CPU cost, but quality is noticeably decreased compared to Ogg Vorbis or even MP3.
+			[b]QOA ([url=https://qoaformat.org/]Quite OK Audio[/url]):[/b] Performs lossy compression on import. CPU cost is slightly higher than IMA-ADPCM, but quality is much higher.
 		</member>
 		<member name="edit/loop_begin" type="int" setter="" getter="" default="0">
 			The begin loop point to use when [member edit/loop_mode] is [b]Forward[/b], [b]Ping-Pong[/b] or [b]Backward[/b]. This is set in seconds after the beginning of the audio file.

editor/import/resource_importer_wav.cpp

@@ -90,7 +90,7 @@ void ResourceImporterWAV::get_import_options(const String &p_path, List<ImportOp
 	r_options->push_back(ImportOption(PropertyInfo(Variant::INT, "edit/loop_mode", PROPERTY_HINT_ENUM, "Detect From WAV,Disabled,Forward,Ping-Pong,Backward", PROPERTY_USAGE_DEFAULT | PROPERTY_USAGE_UPDATE_ALL_IF_MODIFIED), 0));
 	r_options->push_back(ImportOption(PropertyInfo(Variant::INT, "edit/loop_begin"), 0));
 	r_options->push_back(ImportOption(PropertyInfo(Variant::INT, "edit/loop_end"), -1));
-	r_options->push_back(ImportOption(PropertyInfo(Variant::INT, "compress/mode", PROPERTY_HINT_ENUM, "Disabled,RAM (Ima-ADPCM)"), 0));
+	r_options->push_back(ImportOption(PropertyInfo(Variant::INT, "compress/mode", PROPERTY_HINT_ENUM, "Disabled,RAM (Ima-ADPCM),QOA (Quite OK Audio)"), 0));
 }
 
 Error ResourceImporterWAV::import(const String &p_source_file, const String &p_save_path, const HashMap<StringName, Variant> &p_options, List<String> *r_platform_variants, List<String> *r_gen_files, Variant *r_metadata) {
@@ -454,13 +454,13 @@ Error ResourceImporterWAV::import(const String &p_source_file, const String &p_s
 		is16 = false;
 	}
 
-	Vector<uint8_t> dst_data;
+	Vector<uint8_t> pcm_data;
 	AudioStreamWAV::Format dst_format;
 
 	if (compression == 1) {
 		dst_format = AudioStreamWAV::FORMAT_IMA_ADPCM;
 		if (format_channels == 1) {
-			_compress_ima_adpcm(data, dst_data);
+			_compress_ima_adpcm(data, pcm_data);
 		} else {
 			//byte interleave
 			Vector<float> left;
@@ -482,9 +482,9 @@ Error ResourceImporterWAV::import(const String &p_source_file, const String &p_s
 			_compress_ima_adpcm(right, bright);
 
 			int dl = bleft.size();
-			dst_data.resize(dl * 2);
+			pcm_data.resize(dl * 2);
 
-			uint8_t *w = dst_data.ptrw();
+			uint8_t *w = pcm_data.ptrw();
 			const uint8_t *rl = bleft.ptr();
 			const uint8_t *rr = bright.ptr();
 
@@ -496,13 +496,14 @@ Error ResourceImporterWAV::import(const String &p_source_file, const String &p_s
 
 	} else {
 		dst_format = is16 ? AudioStreamWAV::FORMAT_16_BITS : AudioStreamWAV::FORMAT_8_BITS;
-		dst_data.resize(data.size() * (is16 ? 2 : 1));
+		bool enforce16 = is16 || compression == 2;
+		pcm_data.resize(data.size() * (enforce16 ? 2 : 1));
 		{
-			uint8_t *w = dst_data.ptrw();
+			uint8_t *w = pcm_data.ptrw();
 
 			int ds = data.size();
 			for (int i = 0; i < ds; i++) {
-				if (is16) {
+				if (enforce16) {
 					int16_t v = CLAMP(data[i] * 32768, -32768, 32767);
 					encode_uint16(v, &w[i * 2]);
 				} else {
@@ -513,6 +514,23 @@ Error ResourceImporterWAV::import(const String &p_source_file, const String &p_s
 		}
 	}
 
+	Vector<uint8_t> dst_data;
+	if (compression == 2) {
+		dst_format = AudioStreamWAV::FORMAT_QOA;
+		qoa_desc desc = { 0, 0, 0, { { { 0 }, { 0 } } } };
+		uint32_t qoa_len = 0;
+
+		desc.samplerate = rate;
+		desc.samples = frames;
+		desc.channels = format_channels;
+
+		void *encoded = qoa_encode((short *)pcm_data.ptrw(), &desc, &qoa_len);
+		dst_data.resize(qoa_len);
+		memcpy(dst_data.ptrw(), encoded, qoa_len);
+	} else {
+		dst_data = pcm_data;
+	}
+
 	Ref<AudioStreamWAV> sample;
 	sample.instantiate();
 	sample->set_data(dst_data);

scene/resources/audio_stream_wav.cpp

@@ -86,15 +86,15 @@ void AudioStreamPlaybackWAV::seek(double p_time) {
 	offset = uint64_t(p_time * base->mix_rate) << MIX_FRAC_BITS;
 }
 
-template <typename Depth, bool is_stereo, bool is_ima_adpcm>
-void AudioStreamPlaybackWAV::do_resample(const Depth *p_src, AudioFrame *p_dst, int64_t &p_offset, int32_t &p_increment, uint32_t p_amount, IMA_ADPCM_State *p_ima_adpcm) {
+template <typename Depth, bool is_stereo, bool is_ima_adpcm, bool is_qoa>
+void AudioStreamPlaybackWAV::do_resample(const Depth *p_src, AudioFrame *p_dst, int64_t &p_offset, int32_t &p_increment, uint32_t p_amount, IMA_ADPCM_State *p_ima_adpcm, QOA_State *p_qoa) {
 	// this function will be compiled branchless by any decent compiler
 
-	int32_t final, final_r, next, next_r;
+	int32_t final = 0, final_r = 0, next = 0, next_r = 0;
 	while (p_amount) {
 		p_amount--;
 		int64_t pos = p_offset >> MIX_FRAC_BITS;
-		if (is_stereo && !is_ima_adpcm) {
+		if (is_stereo && !is_ima_adpcm && !is_qoa) {
 			pos <<= 1;
 		}
 
@@ -175,32 +175,77 @@ void AudioStreamPlaybackWAV::do_resample(const Depth *p_src, AudioFrame *p_dst,
 			}
 
 		} else {
-			final = p_src[pos];
-			if (is_stereo) {
-				final_r = p_src[pos + 1];
-			}
+			if (is_qoa) {
+				if (pos != p_qoa->cache_pos) { // Prevents triple decoding on lower mix rates.
+					for (int i = 0; i < 2; i++) {
+						// Sign operations prevent triple decoding on backward loops, maxing prevents pop.
+						uint32_t interp_pos = MIN(pos + (i * sign) + (sign < 0), p_qoa->desc->samples - 1);
+						uint32_t new_data_ofs = 8 + interp_pos / QOA_FRAME_LEN * p_qoa->frame_len;
+
+						if (p_qoa->data_ofs != new_data_ofs) {
+							p_qoa->data_ofs = new_data_ofs;
+							const uint8_t *src_ptr = (const uint8_t *)base->data;
+							src_ptr += p_qoa->data_ofs + AudioStreamWAV::DATA_PAD;
+							qoa_decode_frame(src_ptr, p_qoa->frame_len, p_qoa->desc, p_qoa->dec, &p_qoa->dec_len);
+						}
 
-			if constexpr (sizeof(Depth) == 1) { /* conditions will not exist anymore when compiled! */
-				final <<= 8;
+						uint32_t dec_idx = (interp_pos % QOA_FRAME_LEN) * p_qoa->desc->channels;
+
+						if ((sign > 0 && i == 0) || (sign < 0 && i == 1)) {
+							final = p_qoa->dec[dec_idx];
+							p_qoa->cache[0] = final;
+							if (is_stereo) {
+								final_r = p_qoa->dec[dec_idx + 1];
+								p_qoa->cache_r[0] = final_r;
+							}
+						} else {
+							next = p_qoa->dec[dec_idx];
+							p_qoa->cache[1] = next;
+							if (is_stereo) {
+								next_r = p_qoa->dec[dec_idx + 1];
+								p_qoa->cache_r[1] = next_r;
+							}
+						}
+					}
+					p_qoa->cache_pos = pos;
+				} else {
+					final = p_qoa->cache[0];
+					if (is_stereo) {
+						final_r = p_qoa->cache_r[0];
+					}
+
+					next = p_qoa->cache[1];
+					if (is_stereo) {
+						next_r = p_qoa->cache_r[1];
+					}
+				}
+			} else {
+				final = p_src[pos];
 				if (is_stereo) {
-					final_r <<= 8;
+					final_r = p_src[pos + 1];
 				}
-			}
 
-			if (is_stereo) {
-				next = p_src[pos + 2];
-				next_r = p_src[pos + 3];
-			} else {
-				next = p_src[pos + 1];
-			}
+				if constexpr (sizeof(Depth) == 1) { /* conditions will not exist anymore when compiled! */
+					final <<= 8;
+					if (is_stereo) {
+						final_r <<= 8;
+					}
+				}
 
-			if constexpr (sizeof(Depth) == 1) {
-				next <<= 8;
 				if (is_stereo) {
-					next_r <<= 8;
+					next = p_src[pos + 2];
+					next_r = p_src[pos + 3];
+				} else {
+					next = p_src[pos + 1];
 				}
-			}
 
+				if constexpr (sizeof(Depth) == 1) {
+					next <<= 8;
+					if (is_stereo) {
+						next_r <<= 8;
+					}
+				}
+			}
 			int32_t frac = int64_t(p_offset & MIX_FRAC_MASK);
 
 			final = final + ((next - final) * frac >> MIX_FRAC_BITS);
@@ -240,6 +285,9 @@ int AudioStreamPlaybackWAV::mix(AudioFrame *p_buffer, float p_rate_scale, int p_
 		case AudioStreamWAV::FORMAT_IMA_ADPCM:
 			len *= 2;
 			break;
+		case AudioStreamWAV::FORMAT_QOA:
+			len = qoa.desc->samples * qoa.desc->channels;
+			break;
 	}
 
 	if (base->stereo) {
@@ -368,27 +416,34 @@ int AudioStreamPlaybackWAV::mix(AudioFrame *p_buffer, float p_rate_scale, int p_
 		switch (base->format) {
 			case AudioStreamWAV::FORMAT_8_BITS: {
 				if (is_stereo) {
-					do_resample<int8_t, true, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int8_t, true, false, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				} else {
-					do_resample<int8_t, false, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int8_t, false, false, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				}
 			} break;
 			case AudioStreamWAV::FORMAT_16_BITS: {
 				if (is_stereo) {
-					do_resample<int16_t, true, false>((int16_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int16_t, true, false, false>((int16_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				} else {
-					do_resample<int16_t, false, false>((int16_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int16_t, false, false, false>((int16_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				}
 
 			} break;
 			case AudioStreamWAV::FORMAT_IMA_ADPCM: {
 				if (is_stereo) {
-					do_resample<int8_t, true, true>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int8_t, true, true, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				} else {
-					do_resample<int8_t, false, true>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm);
+					do_resample<int8_t, false, true, false>((int8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
 				}
 
 			} break;
+			case AudioStreamWAV::FORMAT_QOA: {
+				if (is_stereo) {
+					do_resample<uint8_t, true, false, true>((uint8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
+				} else {
+					do_resample<uint8_t, false, false, true>((uint8_t *)data, dst_buff, offset, increment, target, ima_adpcm, &qoa);
+				}
+			} break;
 		}
 
 		dst_buff += target;
@@ -412,6 +467,16 @@ void AudioStreamPlaybackWAV::tag_used_streams() {
 
 AudioStreamPlaybackWAV::AudioStreamPlaybackWAV() {}
 
+AudioStreamPlaybackWAV::~AudioStreamPlaybackWAV() {
+	if (qoa.desc) {
+		memfree(qoa.desc);
+	}
+
+	if (qoa.dec) {
+		memfree(qoa.dec);
+	}
+}
+
 /////////////////////
 
 void AudioStreamWAV::set_format(Format p_format) {
@@ -475,6 +540,10 @@ double AudioStreamWAV::get_length() const {
 		case AudioStreamWAV::FORMAT_IMA_ADPCM:
 			len *= 2;
 			break;
+		case AudioStreamWAV::FORMAT_QOA:
+			qoa_desc desc = { 0, 0, 0, { { { 0 }, { 0 } } } };
+			qoa_decode_header((uint8_t *)data + DATA_PAD, QOA_MIN_FILESIZE, &desc);
+			len = desc.samples * desc.channels;
 	}
 
 	if (stereo) {
@@ -526,8 +595,8 @@ Vector<uint8_t> AudioStreamWAV::get_data() const {
 }
 
 Error AudioStreamWAV::save_to_wav(const String &p_path) {
-	if (format == AudioStreamWAV::FORMAT_IMA_ADPCM) {
-		WARN_PRINT("Saving IMA_ADPC samples are not supported yet");
+	if (format == AudioStreamWAV::FORMAT_IMA_ADPCM || format == AudioStreamWAV::FORMAT_QOA) {
+		WARN_PRINT("Saving IMA_ADPCM and QOA samples is not supported yet");
 		return ERR_UNAVAILABLE;
 	}
 
@@ -548,6 +617,7 @@ Error AudioStreamWAV::save_to_wav(const String &p_path) {
 			byte_pr_sample = 1;
 			break;
 		case AudioStreamWAV::FORMAT_16_BITS:
+		case AudioStreamWAV::FORMAT_QOA:
 			byte_pr_sample = 2;
 			break;
 		case AudioStreamWAV::FORMAT_IMA_ADPCM:
@@ -590,6 +660,7 @@ Error AudioStreamWAV::save_to_wav(const String &p_path) {
 			}
 			break;
 		case AudioStreamWAV::FORMAT_16_BITS:
+		case AudioStreamWAV::FORMAT_QOA:
 			for (unsigned int i = 0; i < data_bytes / 2; i++) {
 				uint16_t data_point = decode_uint16(&read_data[i * 2]);
 				file->store_16(data_point);
@@ -607,6 +678,16 @@ Ref<AudioStreamPlayback> AudioStreamWAV::instantiate_playback() {
 	Ref<AudioStreamPlaybackWAV> sample;
 	sample.instantiate();
 	sample->base = Ref<AudioStreamWAV>(this);
+
+	if (format == AudioStreamWAV::FORMAT_QOA) {
+		sample->qoa.desc = (qoa_desc *)memalloc(sizeof(qoa_desc));
+		qoa_decode_header((uint8_t *)data + DATA_PAD, QOA_MIN_FILESIZE, sample->qoa.desc);
+		sample->qoa.frame_len = qoa_max_frame_size(sample->qoa.desc);
+		int samples_len = (sample->qoa.desc->samples > QOA_FRAME_LEN ? QOA_FRAME_LEN : sample->qoa.desc->samples);
+		int alloc_len = sample->qoa.desc->channels * samples_len * sizeof(int16_t);
+		sample->qoa.dec = (int16_t *)memalloc(alloc_len);
+	}
+
 	return sample;
 }
 
@@ -639,7 +720,7 @@ void AudioStreamWAV::_bind_methods() {
 	ClassDB::bind_method(D_METHOD("save_to_wav", "path"), &AudioStreamWAV::save_to_wav);
 
 	ADD_PROPERTY(PropertyInfo(Variant::PACKED_BYTE_ARRAY, "data", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR), "set_data", "get_data");
-	ADD_PROPERTY(PropertyInfo(Variant::INT, "format", PROPERTY_HINT_ENUM, "8-Bit,16-Bit,IMA-ADPCM"), "set_format", "get_format");
+	ADD_PROPERTY(PropertyInfo(Variant::INT, "format", PROPERTY_HINT_ENUM, "8-Bit,16-Bit,IMA-ADPCM,QOA"), "set_format", "get_format");
 	ADD_PROPERTY(PropertyInfo(Variant::INT, "loop_mode", PROPERTY_HINT_ENUM, "Disabled,Forward,Ping-Pong,Backward"), "set_loop_mode", "get_loop_mode");
 	ADD_PROPERTY(PropertyInfo(Variant::INT, "loop_begin"), "set_loop_begin", "get_loop_begin");
 	ADD_PROPERTY(PropertyInfo(Variant::INT, "loop_end"), "set_loop_end", "get_loop_end");
@@ -649,6 +730,7 @@ void AudioStreamWAV::_bind_methods() {
 	BIND_ENUM_CONSTANT(FORMAT_8_BITS);
 	BIND_ENUM_CONSTANT(FORMAT_16_BITS);
 	BIND_ENUM_CONSTANT(FORMAT_IMA_ADPCM);
+	BIND_ENUM_CONSTANT(FORMAT_QOA);
 
 	BIND_ENUM_CONSTANT(LOOP_DISABLED);
 	BIND_ENUM_CONSTANT(LOOP_FORWARD);