Add support for AIFF files and other WAV formats

Author: DeeJayLSP

COPYRIGHT.txt

@@ -231,6 +231,11 @@ Comment: doctest
 Copyright: 2016-2023, Viktor Kirilov
 License: Expat
 
+Files: ./thirdparty/dr_libs/
+Comment: dr_libs
+Copyright: 2024 David Reid
+License: public-domain or Unlicense or Expat
+
 Files: ./thirdparty/embree/
 Comment: Embree
 Copyright: 2009-2021 Intel Corporation

doc/classes/AudioStreamWAV.xml

@@ -1,10 +1,10 @@
 <?xml version="1.0" encoding="UTF-8" ?>
 <class name="AudioStreamWAV" inherits="AudioStream" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="../class.xsd">
 	<brief_description>
-		Stores audio data loaded from WAV files.
+		Stores audio data loaded from WAV or AIFF files.
 	</brief_description>
 	<description>
-		AudioStreamWAV stores sound samples loaded from WAV files. To play the stored sound, use an [AudioStreamPlayer] (for non-positional audio) or [AudioStreamPlayer2D]/[AudioStreamPlayer3D] (for positional audio). The sound can be looped.
+		AudioStreamWAV stores sound samples loaded from WAV or AIFF files. To play the stored sound, use an [AudioStreamPlayer] (for non-positional audio) or [AudioStreamPlayer2D]/[AudioStreamPlayer3D] (for positional audio). The sound can be looped.
 		This class can also be used to store dynamically-generated PCM audio data. See also [AudioStreamGenerator] for procedural audio generation.
 	</description>
 	<tutorials>

doc/classes/ResourceImporterWAV.xml

@@ -1,11 +1,11 @@
 <?xml version="1.0" encoding="UTF-8" ?>
 <class name="ResourceImporterWAV" inherits="ResourceImporter" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="../class.xsd">
 	<brief_description>
-		Imports a WAV audio file for playback.
+		Imports data from a WAV/AIFF audio file for playback.
 	</brief_description>
 	<description>
-		WAV is an uncompressed format, which can provide higher quality compared to Ogg Vorbis and MP3. It also has the lowest CPU cost to decode. This means high numbers of WAV sounds can be played at the same time, even on low-end devices.
-		By default, Redot imports WAV files using the lossy Quite OK Audio compression. You may change this by setting the [member compress/mode] property.
+		WAV/AIFF is an uncompressed format, which can provide higher quality compared to Ogg Vorbis and MP3. It also has the lowest CPU cost to decode. This means a high number of sounds can be played at the same time, even on low-end devices.
+		By default, Redot imports WAV/AIFF files using the lossy Quite OK Audio compression. You may change this by setting the [member compress/mode] property.
 	</description>
 	<tutorials>
 		<link title="Importing audio samples">$DOCS_URL/tutorials/assets_pipeline/importing_audio_samples.html</link>
@@ -25,7 +25,7 @@
 		</member>
 		<member name="edit/loop_mode" type="int" setter="" getter="" default="0">
 			Controls how audio should loop.
-			- [b]Detect From WAV:[/b] Uses loop information from the WAV metadata.
+			- [b]Detect From WAV:[/b] Uses loop information from the WAV metadata (AIFF does not support this).
 			- [b]Disabled:[/b] Don't loop audio, even if the metadata indicates the file playback should loop.
 			- [b]Forward:[/b] Standard audio looping. Plays the audio forward from the beginning to [member edit/loop_end], then returns to [member edit/loop_begin] and repeats.
 			- [b]Ping-Pong:[/b] Plays the audio forward until [member edit/loop_end], then backwards to [member edit/loop_begin], repeating this cycle.

editor/import/resource_importer_wav.cpp

@@ -44,6 +44,10 @@ String ResourceImporterWAV::get_visible_name() const {
 
 void ResourceImporterWAV::get_recognized_extensions(List<String> *p_extensions) const {
 	p_extensions->push_back("wav");
+	p_extensions->push_back("wave");
+	p_extensions->push_back("aif");
+	p_extensions->push_back("aiff");
+	p_extensions->push_back("aifc");
 }
 
 String ResourceImporterWAV::get_save_extension() const {

scene/resources/audio_stream_wav.cpp

@@ -35,6 +35,17 @@
 #include "core/io/file_access_memory.h"
 #include "core/io/marshalls.h"
 
+#define DRWAV_IMPLEMENTATION
+#define DRWAV_NO_STDIO
+#define DR_WAV_LIBSNDFILE_COMPAT
+#define DRWAV_MALLOC(sz) memalloc(sz)
+#define DRWAV_REALLOC(p, sz) memrealloc(p, sz)
+#define DRWAV_FREE(p) \
+	if (p)            \
+	memfree(p)
+
+#include "thirdparty/dr_libs/dr_wav.h"
+
 const float TRIM_DB_LIMIT = -50;
 const int TRIM_FADE_OUT_FRAMES = 500;
 
@@ -599,7 +610,7 @@ Vector<uint8_t> AudioStreamWAV::get_data() const {
 
 Error AudioStreamWAV::save_to_wav(const String &p_path) {
 	if (format == AudioStreamWAV::FORMAT_IMA_ADPCM || format == AudioStreamWAV::FORMAT_QOA) {
-		WARN_PRINT("Saving IMA_ADPCM and QOA samples is not supported yet");
+		WARN_PRINT("Saving IMA ADPCM and QOA samples is not supported.");
 		return ERR_UNAVAILABLE;
 	}
 
@@ -620,11 +631,9 @@ 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:
-			byte_pr_sample = 4;
+		default:
 			break;
 	}
 
@@ -653,8 +662,7 @@ Error AudioStreamWAV::save_to_wav(const String &p_path) {
 	file->store_32(sub_chunk_2_size); //Subchunk2Size
 
 	// Add data
-	Vector<uint8_t> stream_data = get_data();
-	const uint8_t *read_data = stream_data.ptr();
+	const uint8_t *read_data = data.ptr() + DATA_PAD;
 	switch (format) {
 		case AudioStreamWAV::FORMAT_8_BITS:
 			for (unsigned int i = 0; i < data_bytes; i++) {
@@ -663,14 +671,12 @@ 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);
 			}
 			break;
-		case AudioStreamWAV::FORMAT_IMA_ADPCM:
-			//Unimplemented
+		default:
 			break;
 	}
 
@@ -726,220 +732,45 @@ Ref<AudioSample> AudioStreamWAV::generate_sample() const {
 }
 
 Ref<AudioStreamWAV> AudioStreamWAV::load_from_buffer(const Vector<uint8_t> &p_stream_data, const Dictionary &p_options) {
-	// /* STEP 1, READ WAVE FILE */
-
-	Ref<FileAccessMemory> file;
-	file.instantiate();
-	Error err = file->open_custom(p_stream_data.ptr(), p_stream_data.size());
-	ERR_FAIL_COND_V_MSG(err != OK, Ref<AudioStreamWAV>(), "Cannot create memfile for WAV file buffer.");
+	// STEP 1, READ FILE
 
-	/* CHECK RIFF */
-	char riff[5];
-	riff[4] = 0;
-	file->get_buffer((uint8_t *)&riff, 4); //RIFF
-
-	if (riff[0] != 'R' || riff[1] != 'I' || riff[2] != 'F' || riff[3] != 'F') {
-		ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), vformat("Not a WAV file. File should start with 'RIFF', but found '%s', in file of size %d bytes", riff, file->get_length()));
+	drwav wav;
+	if (!drwav_init_memory_with_metadata(&wav, p_stream_data.ptr(), p_stream_data.size(), DRWAV_WITH_METADATA, nullptr)) {
+		ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Audio data is invalid, corrupted or an unsupported format.");
 	}
 
-	/* GET FILESIZE */
-
-	// The file size in header is 8 bytes less than the actual size.
-	// See https://docs.fileformat.com/audio/wav/
-	const int FILE_SIZE_HEADER_OFFSET = 8;
-	uint32_t file_size_header = file->get_32() + FILE_SIZE_HEADER_OFFSET;
-	uint64_t file_size = file->get_length();
-	if (file_size != file_size_header) {
-		WARN_PRINT(vformat("File size %d is %s than the expected size %d.", file_size, file_size > file_size_header ? "larger" : "smaller", file_size_header));
+	if (wav.totalPCMFrameCount > INT32_MAX) {
+		ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Audio data exceeds maximum supported size of 2,147,483,647 frames.");
 	}
 
-	/* CHECK WAVE */
+	int format_bits = wav.bitsPerSample;
+	int format_channels = wav.channels;
+	int format_freq = wav.sampleRate;
+	int frames = wav.totalPCMFrameCount;
 
-	char wave[5];
-	wave[4] = 0;
-	file->get_buffer((uint8_t *)&wave, 4); //WAVE
-
-	if (wave[0] != 'W' || wave[1] != 'A' || wave[2] != 'V' || wave[3] != 'E') {
-		ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), vformat("Not a WAV file. Header should contain 'WAVE', but found '%s', in file of size %d bytes", wave, file->get_length()));
-	}
-
-	// Let users override potential loop points from the WAV.
-	// We parse the WAV loop points only with "Detect From WAV" (0).
 	int import_loop_mode = p_options["edit/loop_mode"];
 
-	int format_bits = 0;
-	int format_channels = 0;
-
-	AudioStreamWAV::LoopMode loop_mode = AudioStreamWAV::LOOP_DISABLED;
-	uint16_t compression_code = 1;
-	bool format_found = false;
-	bool data_found = false;
-	int format_freq = 0;
 	int loop_begin = 0;
 	int loop_end = 0;
-	int frames = 0;
-
-	Vector<float> data;
-
-	while (!file->eof_reached()) {
-		/* chunk */
-		char chunk_id[4];
-		file->get_buffer((uint8_t *)&chunk_id, 4); //RIFF
-
-		/* chunk size */
-		uint32_t chunksize = file->get_32();
-		uint32_t file_pos = file->get_position(); //save file pos, so we can skip to next chunk safely
-
-		if (file->eof_reached()) {
-			//ERR_PRINT("EOF REACH");
-			break;
-		}
-
-		if (chunk_id[0] == 'f' && chunk_id[1] == 'm' && chunk_id[2] == 't' && chunk_id[3] == ' ' && !format_found) {
-			/* IS FORMAT CHUNK */
-
-			//Issue: #7755 : Not a bug - usage of other formats (format codes) are unsupported in current importer version.
-			//Consider revision for engine version 3.0
-			compression_code = file->get_16();
-			if (compression_code != 1 && compression_code != 3) {
-				ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Format not supported for WAVE file (not PCM). Save WAVE files as uncompressed PCM or IEEE float instead.");
-			}
-
-			format_channels = file->get_16();
-			if (format_channels != 1 && format_channels != 2) {
-				ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Format not supported for WAVE file (not stereo or mono).");
-			}
-
-			format_freq = file->get_32(); //sampling rate
-
-			file->get_32(); // average bits/second (unused)
-			file->get_16(); // block align (unused)
-			format_bits = file->get_16(); // bits per sample
-
-			if (format_bits % 8 || format_bits == 0) {
-				ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Invalid amount of bits in the sample (should be one of 8, 16, 24 or 32).");
-			}
-
-			if (compression_code == 3 && format_bits % 32) {
-				ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Invalid amount of bits in the IEEE float sample (should be 32 or 64).");
-			}
-
-			/* Don't need anything else, continue */
-			format_found = true;
-		}
-
-		if (chunk_id[0] == 'd' && chunk_id[1] == 'a' && chunk_id[2] == 't' && chunk_id[3] == 'a' && !data_found) {
-			/* IS DATA CHUNK */
-			data_found = true;
-
-			if (!format_found) {
-				ERR_PRINT("'data' chunk before 'format' chunk found.");
+	AudioStreamWAV::LoopMode loop_mode = AudioStreamWAV::LOOP_DISABLED;
+	if (import_loop_mode == 0) {
+		for (uint32_t meta = 0; meta < wav.metadataCount; ++meta) {
+			drwav_metadata md = wav.pMetadata[meta];
+			if (md.type == drwav_metadata_type_smpl && md.data.smpl.sampleLoopCount > 0) {
+				drwav_smpl_loop loop = md.data.smpl.pLoops[0];
+				loop_mode = (AudioStreamWAV::LoopMode)(loop.type + 1);
+				loop_begin = loop.firstSampleByteOffset;
+				loop_end = loop.lastSampleByteOffset;
 				break;
 			}
-
-			uint64_t remaining_bytes = file_size - file_pos;
-			frames = chunksize;
-			if (remaining_bytes < chunksize) {
-				WARN_PRINT("Data chunk size is smaller than expected. Proceeding with actual data size.");
-				frames = remaining_bytes;
-			}
-
-			ERR_FAIL_COND_V(format_channels == 0, Ref<AudioStreamWAV>());
-			frames /= format_channels;
-			frames /= (format_bits >> 3);
-
-			/*print_line("chunksize: "+itos(chunksize));
-			print_line("channels: "+itos(format_channels));
-			print_line("bits: "+itos(format_bits));
-			*/
-
-			data.resize(frames * format_channels);
-
-			if (compression_code == 1) {
-				if (format_bits == 8) {
-					for (int i = 0; i < frames * format_channels; i++) {
-						// 8 bit samples are UNSIGNED
-
-						data.write[i] = int8_t(file->get_8() - 128) / 128.f;
-					}
-				} else if (format_bits == 16) {
-					for (int i = 0; i < frames * format_channels; i++) {
-						//16 bit SIGNED
-
-						data.write[i] = int16_t(file->get_16()) / 32768.f;
-					}
-				} else {
-					for (int i = 0; i < frames * format_channels; i++) {
-						//16+ bits samples are SIGNED
-						// if sample is > 16 bits, just read extra bytes
-
-						uint32_t s = 0;
-						for (int b = 0; b < (format_bits >> 3); b++) {
-							s |= ((uint32_t)file->get_8()) << (b * 8);
-						}
-						s <<= (32 - format_bits);
-
-						data.write[i] = (int32_t(s) >> 16) / 32768.f;
-					}
-				}
-			} else if (compression_code == 3) {
-				if (format_bits == 32) {
-					for (int i = 0; i < frames * format_channels; i++) {
-						//32 bit IEEE Float
-
-						data.write[i] = file->get_float();
-					}
-				} else if (format_bits == 64) {
-					for (int i = 0; i < frames * format_channels; i++) {
-						//64 bit IEEE Float
-
-						data.write[i] = file->get_double();
-					}
-				}
-			}
-
-			// This is commented out due to some weird edge case seemingly in FileAccessMemory, doesn't seem to have any side effects though.
-			// if (file->eof_reached()) {
-			// 	ERR_FAIL_V_MSG(Ref<AudioStreamWAV>(), "Premature end of file.");
-			// }
 		}
+	}
 
-		if (import_loop_mode == 0 && chunk_id[0] == 's' && chunk_id[1] == 'm' && chunk_id[2] == 'p' && chunk_id[3] == 'l') {
-			// Loop point info!
-
-			/**
-			 *	Consider exploring next document:
-			 *		http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/Docs/RIFFNEW.pdf
-			 *	Especially on page:
-			 *		16 - 17
-			 *	Timestamp:
-			 *		22:38 06.07.2017 GMT
-			 **/
-
-			for (int i = 0; i < 10; i++) {
-				file->get_32(); // i wish to know why should i do this... no doc!
-			}
+	Vector<float> data;
+	data.resize(frames * format_channels);
+	drwav_read_pcm_frames_f32(&wav, frames, data.ptrw());
 
-			// only read 0x00 (loop forward), 0x01 (loop ping-pong) and 0x02 (loop backward)
-			// Skip anything else because it's not supported, reserved for future uses or sampler specific
-			// from https://sites.google.com/site/musicgapi/technical-documents/wav-file-format#smpl (loop type values table)
-			int loop_type = file->get_32();
-			if (loop_type == 0x00 || loop_type == 0x01 || loop_type == 0x02) {
-				if (loop_type == 0x00) {
-					loop_mode = AudioStreamWAV::LOOP_FORWARD;
-				} else if (loop_type == 0x01) {
-					loop_mode = AudioStreamWAV::LOOP_PINGPONG;
-				} else if (loop_type == 0x02) {
-					loop_mode = AudioStreamWAV::LOOP_BACKWARD;
-				}
-				loop_begin = file->get_32();
-				loop_end = file->get_32();
-			}
-		}
-		// Move to the start of the next chunk. Note that RIFF requires a padding byte for odd
-		// chunk sizes.
-		file->seek(file_pos + chunksize + (chunksize & 1));
-	}
+	drwav_uninit(&wav);
 
 	// STEP 2, APPLY CONVERSIONS
 

thirdparty/README.md

@@ -186,6 +186,18 @@ Files extracted from upstream source:
 - `LICENSE.txt`
 
 
+## dr_libs
+
+- Upstream: https://github.com/mackron/dr_libs
+- Version: git (da35f9d6c7374a95353fd1df1d394d44ab66cf01, 2024)
+- License: Public Domain or Unlicense or MIT
+
+Files extracted from upstream source:
+
+- `dr_wav.h`
+- `LICENSE`
+
+
 ## embree
 
 - Upstream: https://github.com/embree/embree