From 23432cb596c3265fd46cc19feb0560456d9ee512 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Sun, 27 Jul 2025 13:57:48 +0700 Subject: [PATCH 01/11] Remove funkin.vis --- libs.xml | 2 -- 1 file changed, 2 deletions(-) diff --git a/libs.xml b/libs.xml index 403f646d..6b16246e 100644 --- a/libs.xml +++ b/libs.xml @@ -21,8 +21,6 @@ - - From c6c7cdd32bc24c0e3300a17116253502bc559f49 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Sun, 27 Jul 2025 18:29:24 +0700 Subject: [PATCH 02/11] Fix & Change a bit of AudioAnalyzer base functions --- source/funkin/backend/utils/AudioAnalyzer.hx | 104 +++++++++++-------- 1 file changed, 58 insertions(+), 46 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 0ad9af3d..90d47d38 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -18,7 +18,11 @@ class AudioAnalyzer { var __wordSize:Int; var __bitUnsignedSize:Int; var __bitSize:Int; + var __min:Array = []; + var __max:Array = []; + //var __ln10:Float = 2.3025850929940456; + //var __ln2:Float = 0.6931471805599453; #if lime_vorbis var __vorbis:VorbisFile; #end @@ -28,36 +32,47 @@ class AudioAnalyzer { __check(); } + public function getLevels(levels:Array, barCount:Int, duration:Float):Float { + __check(); + // TODO: implement FFT + return 0; + } + inline function __check() if (sound.buffer != buffer) { __bitSize = 1 << ((buffer = sound.buffer).bitsPerSample - 1); __bitUnsignedSize = 1 << buffer.bitsPerSample; __toBits = buffer.sampleRate / 1000 * buffer.channels * (__wordSize = buffer.bitsPerSample >> 3); __vorbis = null; + __min.resize(buffer.channels); + __max.resize(buffer.channels); } - public function analyze(startPos:Float, endPos:Float, ?outSeperate:Array):Float { + /** + * Returns a peak of an attached FlxSound from startPos to endPos in milliseconds. + * @param startPos Start Position of the FlxSound in milliseconds. + * @param endPos End Position of the FlxSound in milliseconds. + * @param outMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional) + * @param outMax The output maximum value from the analyzer, indices is in channels (Optional) + * @return Output Amplitude value + **/ + public function analyze(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float { __check(); - if (buffer.data != null) return __analyze(startPos, endPos, outSeperate); + if (buffer.data != null) return __analyze(startPos, endPos, outMin, outMax); #if lime_vorbis - else if (__prepareVorbis()) return __analyzeVorbis(startPos, endPos, outSeperate); + else if (__prepareVorbis()) return __analyzeVorbis(startPos, endPos, outMin, outMax); #end return 0; } - inline function __analyze(startPos:Float, endPos:Float, ?outSeperate:Array):Float { - var pos = Math.floor(startPos * __toBits), end = Math.floor(endPos * __toBits), buf = buffer.data #if !js .buffer #end; - var max = 0, b = 0, w = 0, c = 0, useSeperate = outSeperate != null; - if (useSeperate) { - while (c < outSeperate.length) outSeperate[c++] *= __bitSize; - while (c++ < buffer.channels) outSeperate.push(0); - c = Math.floor((pos % (__wordSize * buffer.channels)) / __wordSize); - } + function __analyze(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float { + __prepareAmplitude(); + var pos = Math.floor(startPos * __toBits), end = Math.floor(endPos * __toBits), buf = buffer.data #if !js .buffer #end; + var c = Math.floor((pos % (__wordSize * buffer.channels)) / __wordSize), b = 0, w = 0; pos -= pos % __wordSize; end -= end % __wordSize; while (pos < end) { - // 8-bit audio data is unsigned (0 is 128, 128 is 256, -128 is 0) if (__wordSize == 1) b = #if js buf[pos++] #else buf.get(pos++) #end - __bitSize; else { while (w < buffer.bitsPerSample) { @@ -67,15 +82,27 @@ class AudioAnalyzer { } if (b > __bitSize) b -= __bitUnsignedSize; } - if (max < b) max = b; - if (useSeperate) { - if (outSeperate[c] < b) outSeperate[c] = b; - if (++c > buffer.channels) c = 0; - } + if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; + if (++c > buffer.channels) c = 0; w = b = 0; } - return max / __bitSize; + return __getAmplitude(outMin, outMax); + } + + inline function __prepareAmplitude() for (i in 0...buffer.channels) __min[i] = __max[i] = 0; + inline function __getAmplitude(?outMin:Array, ?outMax:Array):Float { + var min = 0, max = 0, useOutput = outMin != null && outMax != null, v:Float; + for (i in 0...buffer.channels) { + if (__min[i] > min) min = __min[i]; + if (__max[i] > max) max = __max[i]; + if (useOutput) { + if (outMin[i] < (v = __min[i] / __bitSize)) outMin[i] = v; + if (outMax[i] < (v = __max[i] / __bitSize)) outMax[i] = v; + } + } + + return (max + min) / __bitSize; } #if lime_vorbis // As far i know, only native supports vorbis @@ -91,17 +118,13 @@ class AudioAnalyzer { return false; } - inline function __analyzeVorbis(startPos:Float, endPos:Float, ?outSeperate:Array):Float @:privateAccess { + function __analyzeVorbis(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float @:privateAccess { var n = Math.floor((endPos - startPos) * __toBits); if ((n -= n % __wordSize) < __wordSize) return 0; - var pos = 0, max = 0, b = 0, w = 0, c = 0, useSeperate = outSeperate != null; - if (useSeperate) { - while (c < outSeperate.length) outSeperate[c++] *= __bitSize; - while (c++ < buffer.channels) outSeperate.push(0); - } + var pos = 0, c = 0, b = 0, w = 0; + __prepareAmplitude(); - // IT SHOULD BE ALWAYS BACKEND.STREAMED IF THIS GETS CALLED. var backend = sound._source != null ? sound._source.__backend : null; if (backend != null && backend.streamTimer != null) { var i = backend.bufferSizes.length - backend.queuedBuffers; @@ -109,29 +132,21 @@ class AudioAnalyzer { if (startPos >= time && startPos < backend.bufferTimes[backend.bufferSizes.length - 1] * 1000) { var buf = backend.bufferDatas[i].buffer, size = backend.bufferSizes[i]; - pos = Math.floor((startPos - time) * __toBits); - c = Math.floor((pos % (__wordSize * buffer.channels)) / __wordSize); + c = Math.floor(((pos = Math.floor((startPos - time) * __toBits)) % (__wordSize * buffer.channels)) / __wordSize); pos -= pos % __wordSize; while (n > 0) { - if (__wordSize == 1) { - b = buf.get(pos++) - __bitSize; - n--; - } + if (__wordSize == 1) b = buf.get(pos++) - __bitSize; else { while (w < buffer.bitsPerSample) { b |= buf.get(pos) << w; w += 8; pos++; - n--; } if (b > __bitSize) b -= __bitUnsignedSize; } - if (max < b) max = b; - if (useSeperate) { - if (outSeperate[c] < b) outSeperate[c] = b; - if (++c > buffer.channels) c = 0; - } + if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; + if (++c > buffer.channels) c = 0; if (pos >= size) { if (++i >= backend.bufferDatas.length) break; size = backend.bufferSizes[i]; @@ -139,20 +154,20 @@ class AudioAnalyzer { pos = 0; } w = b = 0; + n -= __wordSize; } + if (n == 0) return __getAmplitude(outMin, outMax); startPos = pos / __toBits + time; - // Cannot inline a not final return + pos = c = 0; } } - if (n == 0) return max / __bitSize; if (Math.abs(startPos - __vorbis.timeTell() * 1000) > 4) { if (startPos < 1) __vorbis.rawSeek(0); else __vorbis.timeSeek(startPos / 1000); } - pos = c = 0; var isBigEndian = lime.system.System.endianness == lime.system.Endian.BIG_ENDIAN, result; while (n > 0) { n -= (result = __vorbis.read(__buffer, 0, n < __bufferSize ? n : __bufferSize, isBigEndian, __wordSize, true)); @@ -170,17 +185,14 @@ class AudioAnalyzer { } if (b > __bitSize) b -= __bitUnsignedSize; } - if (max < b) max = b; - if (useSeperate) { - if (outSeperate[c] < b) outSeperate[c] = b; - if (++c > buffer.channels) c = 0; - } + if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; + if (++c > buffer.channels) c = 0; w = b = 0; } pos = 0; } - return max / __bitSize; + return __getAmplitude(outMin, outMax); } #end } \ No newline at end of file From 92ee5a280ace81b7c272e45c6495ba692400b2fe Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Sun, 27 Jul 2025 18:49:23 +0700 Subject: [PATCH 03/11] fix SoundChannel peaks & give AudioAnalyzer desc --- source/funkin/backend/utils/AudioAnalyzer.hx | 17 ++++++++++++++--- source/openfl/media/SoundChannel.hx | 14 ++++++++------ 2 files changed, 22 insertions(+), 9 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 90d47d38..34bd3277 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -9,8 +9,13 @@ import lime.media.vorbis.VorbisFile; import lime.utils.ArrayBuffer; #end -// ORIGINAL CODES FROM YOSH & LUNAR https://github.com/CodenameCrew/YoshiCrafterEngine/blob/main/source/WaveformSprite.hx -// REWRITTEN BY RALTYRO +/** + * An utility that analyze FlxSounds, + * can be used to make waveform or real-time audio visualizer. + * + * FlxSound.amplitude does work in CNE so if any case if your only checking for peak of current + * time, use that instead. +**/ class AudioAnalyzer { public var sound:FlxSound; public var buffer:AudioBuffer; @@ -27,14 +32,20 @@ class AudioAnalyzer { var __vorbis:VorbisFile; #end + /** + * Creates an analyzer for specified FlxSound + * @param sound An FlxSound to analyze. + **/ public function new(sound:FlxSound) { this.sound = sound; __check(); } + /** + * TODO: IMPLEMENT FFT + **/ public function getLevels(levels:Array, barCount:Int, duration:Float):Float { __check(); - // TODO: implement FFT return 0; } diff --git a/source/openfl/media/SoundChannel.hx b/source/openfl/media/SoundChannel.hx index 6ccbf8ac..ae3ddf72 100644 --- a/source/openfl/media/SoundChannel.hx +++ b/source/openfl/media/SoundChannel.hx @@ -152,9 +152,7 @@ import lime.media.openal.AL; @:noCompletion private function __updatePeaks(time:Float):Bool { if (Math.abs(time - __lastPeakTime) < 8) return false; - __lastPeakTime = time; - __leftPeak = __rightPeak = 0; #if !macro if (!__isValid) return false; @@ -162,6 +160,7 @@ import lime.media.openal.AL; var buffer = __source.buffer; var wordSize = buffer.bitsPerSample >> 3, bitUnsignedSize = 1 << buffer.bitsPerSample, bitSize = 1 << (buffer.bitsPerSample - 1); var pos = Math.floor(time * buffer.sampleRate / 1000 * buffer.channels * wordSize), size = 0, buf; + var leftMin = 0, leftMax = 0, rightMin = 0, rightMax = 0; #if lime_cffi var backend = __source.__backend, i = 0; @@ -189,7 +188,10 @@ import lime.media.openal.AL; } if (b > bitSize) b -= bitUnsignedSize; } - ((c % 2 == 0) ? (if (__leftPeak < b) __leftPeak = b) : (if (__rightPeak < b) __rightPeak = b)); + if (c % 2 == 0) { + if (leftMax < b) leftMax = b; else if (leftMin < (b = -b)) leftMin = b; + } + else if (rightMax < b) rightMax = b; else if (rightMin < (b = -b)) rightMin = b; if (pos >= size) #if lime_cffi { if (!backend.streamed || ++i >= backend.bufferDatas.length) break; @@ -206,10 +208,10 @@ import lime.media.openal.AL; w = b = 0; } - if (buffer.channels == 1) __rightPeak = __leftPeak / bitSize; + if (buffer.channels == 1) __rightPeak = (__leftPeak = (leftMax + leftMin) / bitSize); else { - __leftPeak = __leftPeak / bitSize; - __rightPeak = __rightPeak / bitSize; + __leftPeak = (leftMax + leftMin) / bitSize; + __rightPeak = (rightMax + rightMin) / bitSize; } #end From 16f5d5010f20ff99731fb9e750b1f23d1a6dea71 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Sun, 27 Jul 2025 18:50:35 +0700 Subject: [PATCH 04/11] Completely remove funkin.vis --- project.xml | 5 ----- 1 file changed, 5 deletions(-) diff --git a/project.xml b/project.xml index df54e556..97c07c70 100644 --- a/project.xml +++ b/project.xml @@ -153,9 +153,6 @@ - - - @@ -196,8 +193,6 @@ - - From 2724c8b236cd39b720a33a80a0f93e7fe5237957 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Wed, 30 Jul 2025 20:48:42 +0700 Subject: [PATCH 05/11] Cleanup AudioAnalyzer --- source/funkin/backend/utils/AudioAnalyzer.hx | 298 +++++++++++-------- source/openfl/media/SoundChannel.hx | 38 +-- 2 files changed, 185 insertions(+), 151 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 34bd3277..48683859 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -2,11 +2,12 @@ package funkin.backend.utils; import flixel.sound.FlxSound; import lime.media.AudioBuffer; +import lime.utils.ArrayBufferView.ArrayBufferIO; +import lime.utils.ArrayBuffer; -#if lime_vorbis +#if (lime_cffi && lime_vorbis) import lime.media.vorbis.Vorbis; import lime.media.vorbis.VorbisFile; -import lime.utils.ArrayBuffer; #end /** @@ -17,193 +18,238 @@ import lime.utils.ArrayBuffer; * time, use that instead. **/ class AudioAnalyzer { + public static function getByte(buf:ArrayBuffer, pos:Int, wordSize:Int):Int { + if (wordSize == 2) return ArrayBufferIO.getInt16(buf, pos); + else if (wordSize == 3) { + var b = ArrayBufferIO.getUint16(buf, pos) | (buf.get(pos + 2) << 16); + if (b > 8388608) return b - 16777216; + else return b; + } + else if (wordSize == 4) return ArrayBufferIO.getInt32(buf, pos); + else return ArrayBufferIO.getUint8(buf, pos) - 128; + } + public var sound:FlxSound; public var buffer:AudioBuffer; + public var fftSize:Int; + + public var byteSize:Int; + var __toBits:Float; var __wordSize:Int; - var __bitUnsignedSize:Int; - var __bitSize:Int; + var __sampleSize:Int; + + #if (lime_cffi && lime_vorbis) + var __vorbis:VorbisFile; + var __buffer:ArrayBuffer; + var __bufferSize:Int; + #end + + // analyze var __min:Array = []; var __max:Array = []; - - //var __ln10:Float = 2.3025850929940456; - //var __ln2:Float = 0.6931471805599453; - #if lime_vorbis - var __vorbis:VorbisFile; - #end + var __minByte:Int; + var __maxByte:Int; /** * Creates an analyzer for specified FlxSound * @param sound An FlxSound to analyze. **/ - public function new(sound:FlxSound) { + public function new(sound:FlxSound, fftSize = 1024) { this.sound = sound; + this.fftSize = fftSize; __check(); } + function __check() if (sound.buffer != buffer) { + byteSize = 1 << ((buffer = sound.buffer).bitsPerSample - 1); + + __toBits = buffer.sampleRate / 1000 * (__sampleSize = buffer.channels * (__wordSize = buffer.bitsPerSample >> 3)); + __min.resize(buffer.channels); + __max.resize(buffer.channels); + + #if (lime_cffi && lime_vorbis) __vorbis = null; #end + } + /** * TODO: IMPLEMENT FFT **/ public function getLevels(levels:Array, barCount:Int, duration:Float):Float { - __check(); return 0; } - inline function __check() if (sound.buffer != buffer) { - __bitSize = 1 << ((buffer = sound.buffer).bitsPerSample - 1); - __bitUnsignedSize = 1 << buffer.bitsPerSample; - __toBits = buffer.sampleRate / 1000 * buffer.channels * (__wordSize = buffer.bitsPerSample >> 3); - __vorbis = null; - __min.resize(buffer.channels); - __max.resize(buffer.channels); - } - /** * Returns a peak of an attached FlxSound from startPos to endPos in milliseconds. * @param startPos Start Position of the FlxSound in milliseconds. * @param endPos End Position of the FlxSound in milliseconds. - * @param outMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional) + * @param outOrOutMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional, if outMax doesn't get passed in, it will be [min, max] with all channels combined instead) * @param outMax The output maximum value from the analyzer, indices is in channels (Optional) * @return Output Amplitude value **/ - public function analyze(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float { - __check(); - if (buffer.data != null) return __analyze(startPos, endPos, outMin, outMax); - #if lime_vorbis - else if (__prepareVorbis()) return __analyzeVorbis(startPos, endPos, outMin, outMax); - #end - return 0; + public function analyze(startPos:Float, endPos:Float, ?outOrOutMin:Array, ?outMax:Array):Float { + __minByte = __maxByte = 0; + + if (outOrOutMin != null) { + var f:Float; + if (outMax != null) { + for (i in 0...buffer.channels) __min[i] = __max[i] = 0; + + read(startPos, endPos, __analyzeCallback); + + for (i in 0...buffer.channels) { + if (outOrOutMin[i] < (f = __min[i] / byteSize)) outOrOutMin[i] = f; + if (outMax[i] < (f = __max[i] / byteSize)) outMax[i] = f; + } + } + else { + read(startPos, endPos, __analyzeCallbackSimple); + + outOrOutMin.resize(2); + if (outOrOutMin[0] < (f = __minByte / byteSize)) outOrOutMin[0] = f; + if (outOrOutMin[1] < (f = __maxByte / byteSize)) outOrOutMin[1] = f; + } + } + else + read(startPos, endPos, __analyzeCallbackSimple); + + return (__maxByte + __minByte) / byteSize; } - function __analyze(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float { - __prepareAmplitude(); + function __analyzeCallback(b:Int, c:Int):Void + ((b > __max[c]) ? (if ((__max[c] = b) > __maxByte) (__maxByte = b)) : (if (-b > __min[c]) (if ((__min[c] = -b) > __minByte) (__minByte = __min[c])))); - var pos = Math.floor(startPos * __toBits), end = Math.floor(endPos * __toBits), buf = buffer.data #if !js .buffer #end; - var c = Math.floor((pos % (__wordSize * buffer.channels)) / __wordSize), b = 0, w = 0; - pos -= pos % __wordSize; - end -= end % __wordSize; + function __analyzeCallbackSimple(b:Int, c:Int):Void + ((b > __maxByte) ? (__maxByte = b) : (if (-b > __minByte) (__minByte = -b))); + + public function read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { + __check(); + if (buffer.data != null) __read(startPos, endPos, callback); + #if lime_cffi + else if (__canReadStream() && (startPos += __readStream(startPos, endPos, callback)) >= endPos) return; + #if lime_vorbis + else if (__prepareDecoder()) __readDecoder(startPos, endPos, callback); + #end + #end + } + + inline function __read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { + var pos = Math.floor(startPos * __toBits), end = Math.floor(endPos * __toBits), c = 0; + pos -= pos % __sampleSize; + end -= end % __sampleSize; while (pos < end) { - if (__wordSize == 1) b = #if js buf[pos++] #else buf.get(pos++) #end - __bitSize; - else { - while (w < buffer.bitsPerSample) { - b |= #if js buf[pos] #else buf.get(pos) #end << w; - w += 8; - pos++; - } - if (b > __bitSize) b -= __bitUnsignedSize; - } - if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; + callback(getByte(buffer.data.buffer, pos, __wordSize), c); if (++c > buffer.channels) c = 0; - w = b = 0; + pos += __wordSize; } - - return __getAmplitude(outMin, outMax); } - inline function __prepareAmplitude() for (i in 0...buffer.channels) __min[i] = __max[i] = 0; - inline function __getAmplitude(?outMin:Array, ?outMax:Array):Float { - var min = 0, max = 0, useOutput = outMin != null && outMax != null, v:Float; - for (i in 0...buffer.channels) { - if (__min[i] > min) min = __min[i]; - if (__max[i] > max) max = __max[i]; - if (useOutput) { - if (outMin[i] < (v = __min[i] / __bitSize)) outMin[i] = v; - if (outMax[i] < (v = __max[i] / __bitSize)) outMax[i] = v; + #if lime_cffi + inline function __canReadStream():Bool + @:privateAccess return sound._source != null && sound._source.__backend != null && sound._source.__backend.streamTimer != null; + + inline function __readStream(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback):Float @:privateAccess { + var backend = sound._source.__backend; + var i = backend.bufferSizes.length - backend.queuedBuffers; + var time = backend.bufferTimes[i] * 1000; + + var n = Math.floor((endPos - startPos) * __toBits); + if (startPos >= time && startPos < backend.bufferTimes[backend.bufferSizes.length - 1] * 1000) { + var pos = Math.floor((startPos - time) * __toBits), buf = backend.bufferDatas[i].buffer, size = backend.bufferSizes[i], c = 0; + pos -= pos % __sampleSize; + + while (n > 0) { + callback(getByte(buf, pos, __wordSize), c); + if (++c > buffer.channels) c = 0; + if ((pos += __wordSize) >= size) { + if (++i >= backend.bufferDatas.length) break; + buf = backend.bufferDatas[i].buffer; + size = backend.bufferSizes[i]; + pos = 0; + } + n -= __wordSize; } } - return (max + min) / __bitSize; + return endPos - (n / __toBits); } - #if lime_vorbis // As far i know, only native supports vorbis - var __buffer:ArrayBuffer; - var __bufferSize:Int; - function __prepareVorbis():Bool @:privateAccess { + #if lime_vorbis + inline function __prepareDecoder():Bool @:privateAccess { if (buffer.__srcVorbisFile == null) return __vorbis != null; if (__vorbis != null) return true; if ((__vorbis = buffer.__srcVorbisFile.clone()) != null) { // IM HOPING IT HAVE A GC CLOSURE. - __buffer = new ArrayBuffer(__bufferSize = 0x400 * buffer.channels * (buffer.bitsPerSample >> 3)); + __buffer = new ArrayBuffer(__bufferSize = 0x400 * __sampleSize); return true; } return false; } - function __analyzeVorbis(startPos:Float, endPos:Float, ?outMin:Array, ?outMax:Array):Float @:privateAccess { - var n = Math.floor((endPos - startPos) * __toBits); - if ((n -= n % __wordSize) < __wordSize) return 0; - - var pos = 0, c = 0, b = 0, w = 0; - __prepareAmplitude(); - - var backend = sound._source != null ? sound._source.__backend : null; - if (backend != null && backend.streamTimer != null) { - var i = backend.bufferSizes.length - backend.queuedBuffers; - var time = backend.bufferTimes[i] * 1000; - - if (startPos >= time && startPos < backend.bufferTimes[backend.bufferSizes.length - 1] * 1000) { - var buf = backend.bufferDatas[i].buffer, size = backend.bufferSizes[i]; - c = Math.floor(((pos = Math.floor((startPos - time) * __toBits)) % (__wordSize * buffer.channels)) / __wordSize); - pos -= pos % __wordSize; - - while (n > 0) { - if (__wordSize == 1) b = buf.get(pos++) - __bitSize; - else { - while (w < buffer.bitsPerSample) { - b |= buf.get(pos) << w; - w += 8; - pos++; - } - if (b > __bitSize) b -= __bitUnsignedSize; - } - if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; - if (++c > buffer.channels) c = 0; - if (pos >= size) { - if (++i >= backend.bufferDatas.length) break; - size = backend.bufferSizes[i]; - buf = backend.bufferDatas[i].buffer; - pos = 0; - } - w = b = 0; - n -= __wordSize; - } - - if (n == 0) return __getAmplitude(outMin, outMax); - startPos = pos / __toBits + time; - pos = c = 0; - } - } - - if (Math.abs(startPos - __vorbis.timeTell() * 1000) > 4) { + inline function __readDecoder(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { + var time = startPos / 1000; + if (Math.abs(time - __vorbis.timeTell()) > 0.004) { if (startPos < 1) __vorbis.rawSeek(0); - else __vorbis.timeSeek(startPos / 1000); + else __vorbis.timeSeek(time); } var isBigEndian = lime.system.System.endianness == lime.system.Endian.BIG_ENDIAN, result; - while (n > 0) { - n -= (result = __vorbis.read(__buffer, 0, n < __bufferSize ? n : __bufferSize, isBigEndian, __wordSize, true)); + var n = Math.floor((endPos - startPos) * __toBits), pos = 0, c = 0; + n -= n % __sampleSize; + while (n > 0) { + result = __vorbis.read(__buffer, 0, n < __bufferSize ? n : __bufferSize, isBigEndian, __wordSize, true); if (result == Vorbis.HOLE) continue; else if (result <= 0) break; while (pos < result) { - if (__wordSize == 1) b = __buffer.get(pos++) - __bitSize; - else { - while (w < buffer.bitsPerSample) { - b |= __buffer.get(pos) << w; - w += 8; - pos++; - } - if (b > __bitSize) b -= __bitUnsignedSize; - } - if (__max[c] < b) __max[c] = b; else if (__min[c] < (b = -b)) __min[c] = b; + callback(getByte(__buffer, pos, __wordSize), c); if (++c > buffer.channels) c = 0; - w = b = 0; + if ((pos += __wordSize) >= n) break; } pos = 0; + n -= result; } - - return __getAmplitude(outMin, outMax); } #end -} \ No newline at end of file + #end +} + +private typedef AudioAnalyzerCallback = Int->Int->Void; + +/* +abstract Complex(Array) { + public var real(get, never):Float; + inline function get_real() return this[0]; + + public var imag(get, never):Float; + inline function get_imag() return this[1]; + + public inline function new(real:Float, imag:Float) this = [real, imag]; + public static inline function fromReal(real:Float):Complex return new Complex(real, 0); + public static inline function exp(w:Float):Complex return new Complex(Math.cos(w), Math.sin(w)); + + public var angle(get, never):Float; + inline function get_angle() return Math.atan2(imag, real); + + public var magnitude(get, never):Float; + inline function get_magnitude():Float return Math.sqrt(real*real + imag*imag); + + @:op(A + B) + public inline function add(rhs:Complex):Complex return new Complex(real + rhs.real, imag + rhs.imag); + + @:op(A - B) + public inline function sub(rhs:Complex):Complex return new Complex(real - rhs.real, imag - rhs.imag); + + @:op(A * B) + public inline function mult(rhs:Complex):Complex + return new Complex(real*rhs.real - imag*rhs.imag, real*rhs.imag + imag*rhs.real); + + @:op(A / B) + public inline function div(rhs:Complex):Complex { + var m = rhs.magnitude; + return new Complex((real*rhs.real + imag*rhs.imag) / m, (imag*rhs.real - real*rhs.imag) / m); + } + + public inline function conj():Complex return new Complex(real, -imag); +} +*/ \ No newline at end of file diff --git a/source/openfl/media/SoundChannel.hx b/source/openfl/media/SoundChannel.hx index ae3ddf72..4d9323f5 100644 --- a/source/openfl/media/SoundChannel.hx +++ b/source/openfl/media/SoundChannel.hx @@ -27,7 +27,7 @@ import lime.media.openal.AL; @:fileXml('tags="haxe,release"') @:noDebug #end -#if (lime && lime_cffi) +#if lime_cffi @:access(lime._internal.backend.native.NativeAudioSource) @:access(lime.media.AudioSource) #end @@ -158,9 +158,9 @@ import lime.media.openal.AL; if (!__isValid) return false; var buffer = __source.buffer; - var wordSize = buffer.bitsPerSample >> 3, bitUnsignedSize = 1 << buffer.bitsPerSample, bitSize = 1 << (buffer.bitsPerSample - 1); - var pos = Math.floor(time * buffer.sampleRate / 1000 * buffer.channels * wordSize), size = 0, buf; - var leftMin = 0, leftMax = 0, rightMin = 0, rightMax = 0; + var wordSize = buffer.bitsPerSample >> 3, byteSize = 1 << (buffer.bitsPerSample - 1); + var pos = Math.floor(time * buffer.sampleRate / 1000 * buffer.channels * wordSize); + var leftMin = 0, leftMax = 0, rightMin = 0, rightMax = 0, size = 0, buf; #if lime_cffi var backend = __source.__backend, i = 0; @@ -175,25 +175,14 @@ import lime.media.openal.AL; size = #if js buf.byteLength #else buf.length #end; } - var s = Math.floor(Math.min(buffer.sampleRate / 80, 512)), c = Math.floor((pos % (wordSize * buffer.channels)) / wordSize), b = 0, w = 0; - pos -= pos % wordSize; + var s = Math.floor(Math.min(buffer.sampleRate / 80, 512)), c = 0, b; + pos -= pos % (buffer.channels * wordSize); while (s > 0) { - if (wordSize == 1) b = #if js buf[pos++] #else buf.get(pos++) #end - bitSize; - else { - while (w < buffer.bitsPerSample) { - b |= #if js buf[pos] #else buf.get(pos) #end << w; - w += 8; - pos++; - } - if (b > bitSize) b -= bitUnsignedSize; - } - if (c % 2 == 0) { - if (leftMax < b) leftMax = b; else if (leftMin < (b = -b)) leftMin = b; - } - else if (rightMax < b) rightMax = b; else if (rightMin < (b = -b)) rightMin = b; - - if (pos >= size) #if lime_cffi { + b = funkin.backend.utils.AudioAnalyzer.getByte(buf, pos, wordSize); + if (c % 2 == 0) ((b > leftMax) ? (leftMax = b) : (if ((b = -b) > leftMin) (leftMin = b))); + else ((b > rightMax) ? (rightMax = b) : (if ((b = -b) > rightMin) (rightMin = b))); + if ((pos += wordSize) >= size) #if lime_cffi { if (!backend.streamed || ++i >= backend.bufferDatas.length) break; size = backend.bufferSizes[i]; buf = backend.bufferDatas[i].buffer; @@ -205,13 +194,12 @@ import lime.media.openal.AL; c = 0; s--; } - w = b = 0; } - if (buffer.channels == 1) __rightPeak = (__leftPeak = (leftMax + leftMin) / bitSize); + if (buffer.channels == 1) __rightPeak = (__leftPeak = (leftMax + leftMin) / byteSize); else { - __leftPeak = (leftMax + leftMin) / bitSize; - __rightPeak = (rightMax + rightMin) / bitSize; + __leftPeak = (leftMax + leftMin) / byteSize; + __rightPeak = (rightMax + rightMin) / byteSize; } #end From 1f5e048945077ff073daf546d45f46085008c5aa Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Thu, 31 Jul 2025 19:32:21 +0700 Subject: [PATCH 06/11] First attempt on reimplementing FFT --- source/funkin/backend/utils/AudioAnalyzer.hx | 267 ++++++++++++++----- 1 file changed, 193 insertions(+), 74 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 48683859..b05ad285 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -10,6 +10,8 @@ import lime.media.vorbis.Vorbis; import lime.media.vorbis.VorbisFile; #end +typedef AudioAnalyzerCallback = Int->Int->Void; + /** * An utility that analyze FlxSounds, * can be used to make waveform or real-time audio visualizer. @@ -19,21 +21,21 @@ import lime.media.vorbis.VorbisFile; **/ class AudioAnalyzer { public static function getByte(buf:ArrayBuffer, pos:Int, wordSize:Int):Int { - if (wordSize == 2) return ArrayBufferIO.getInt16(buf, pos); + if (wordSize == 2) return inline ArrayBufferIO.getInt16(buf, pos); else if (wordSize == 3) { - var b = ArrayBufferIO.getUint16(buf, pos) | (buf.get(pos + 2) << 16); + var b = inline ArrayBufferIO.getUint16(buf, pos) | (buf.get(pos + 2) << 16); if (b > 8388608) return b - 16777216; else return b; } - else if (wordSize == 4) return ArrayBufferIO.getInt32(buf, pos); - else return ArrayBufferIO.getUint8(buf, pos) - 128; + else if (wordSize == 4) return inline ArrayBufferIO.getInt32(buf, pos); + else return inline ArrayBufferIO.getUint8(buf, pos) - 128; } public var sound:FlxSound; - public var buffer:AudioBuffer; - public var fftSize:Int; + public var buffer(default, null):AudioBuffer; + public var fftSamples(default, set):Int; - public var byteSize:Int; + public var byteSize(default, null):Int; var __toBits:Float; var __wordSize:Int; @@ -51,66 +53,178 @@ class AudioAnalyzer { var __minByte:Int; var __maxByte:Int; + // samples + var __sampleIndex:Int; + var __sampleOutputLength:Int; + var __sampleOutput:Array; + + // fft + var __logN:Int; + var __freqSamples:Array; + var __reverseIndices:Array = []; + var __rootReals:Array = []; + var __rootImags:Array = []; + var __freqReals:Array = []; + var __freqImags:Array = []; + + // levels + var __frequencies:Array; + /** * Creates an analyzer for specified FlxSound * @param sound An FlxSound to analyze. **/ - public function new(sound:FlxSound, fftSize = 1024) { + public function new(sound:FlxSound, fftSamples = 512) { this.sound = sound; - this.fftSize = fftSize; + this.fftSamples = fftSamples; __check(); } function __check() if (sound.buffer != buffer) { byteSize = 1 << ((buffer = sound.buffer).bitsPerSample - 1); + #if (lime_cffi && lime_vorbis) __vorbis = null; #end + __toBits = buffer.sampleRate / 1000 * (__sampleSize = buffer.channels * (__wordSize = buffer.bitsPerSample >> 3)); __min.resize(buffer.channels); __max.resize(buffer.channels); + } - #if (lime_cffi && lime_vorbis) __vorbis = null; #end + inline function set_fftSamples(v:Int):Int { + if (fftSamples == (fftSamples = v)) return v; + + var half = v >> 1, a, c; + __logN = Math.floor(Math.log(v) / Math.log(2)); + __reverseIndices.resize(half); + __rootReals.resize(half); + __rootImags.resize(half); + __freqReals.resize(half); + __freqImags.resize(half); + for (i in 0...half) { + __reverseIndices[i] = __bitReverse(i); + __rootReals[i] = Math.cos(a = 2 * Math.PI * i / v); + __rootImags[i] = Math.sin(a); + } + + return v; + } + + inline function __bitReverse(x:Int):Int { + var y = 0, i = __logN; + while (i > 0) { + y = (y << 1) | (x & 1); + x >>= 1; + i--; + } + return y; + } + + public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, minDb = -70.0, maxDb = -20.0):Array { + __frequencies = getFrequencies(startPos, __frequencies); + + if (levels == null) levels = []; + levels.resize(barCount); + + var x1 = 0.5, i1 = 1, n = fftSamples >> 1; + var x2, i2, v; + for (i in 0...barCount) { + i2 = Math.floor(x2 = Math.pow(n, (i + 1) / barCount) - 0.5); + + if (i2 < i1) v = __frequencies[i2] * (x2 - x1); + else { + v = __frequencies[i1 - 1] * (i1 - x1); + while (i1 < i2) v += __frequencies[i1++]; + if (i2 < n) v += __frequencies[i2] * (x2 - i2); + } + + i1 = Math.ceil(x1 = x2); + + v = 20 * Math.log(v * barCount / 12) / 2.302585092994046/*log(10)*/ / (maxDb - minDb); + if (delta > 0 && levels[i] > v) levels[i] -= (levels[i] - v) * delta; + else levels[i] = v; + } + + return levels; } /** - * TODO: IMPLEMENT FFT + * Gets frequencies from an attached FlxSound from startPos with samples + * @return Output of frequencies ranging from 1 to fftSamples/2. **/ - public function getLevels(levels:Array, barCount:Int, duration:Float):Float { - return 0; + public function getFrequencies(startPos:Float, ?frequencies:Array):Array { + __freqSamples = getSamples(startPos, fftSamples, true, __freqSamples); + + var inv = fftSamples >> 1; + for (i in 0...inv) { + __freqReals[__reverseIndices[i]] = __freqSamples[i]; + __freqImags[i] = 0; + } + + var h = 1, g = 0, b = 0, r = 0; + var tr, ti, i1, i2; + while (inv > 0) { + h <<= 1; + while (g < fftSamples) { + while (b < h) { + i2 = (i1 = g + b) + h; + tr = __rootReals[r] * __freqReals[i2] - __rootImags[r] * __freqImags[i2]; + ti = __rootReals[r] * __freqImags[i2] + __rootImags[r] * __freqReals[i2]; + + __rootReals[i2] = __freqReals[i1] - tr; + __rootImags[i2] = __freqImags[i1] - ti; + __freqReals[i1] += tr; + __freqImags[i1] += ti; + + b++; + r += inv; + } + b = r = 0; + g += h; + } + g = 0; + + inv >>= 1; + } + + if (frequencies == null) frequencies = []; + frequencies.resize(inv = fftSamples >> 1); + + var mag; + for (i in 0...inv) frequencies[i] = (mag = Math.sqrt(__freqReals[i] * __freqReals[i] + __freqImags[i] * __freqImags[i])) < 0.001 ? 0 : mag; + + return frequencies; } /** - * Returns a peak of an attached FlxSound from startPos to endPos in milliseconds. + * Analyzes an attached FlxSound from startPos to endPos in milliseconds to get the amplitudes. * @param startPos Start Position of the FlxSound in milliseconds. * @param endPos End Position of the FlxSound in milliseconds. - * @param outOrOutMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional, if outMax doesn't get passed in, it will be [min, max] with all channels combined instead) - * @param outMax The output maximum value from the analyzer, indices is in channels (Optional) - * @return Output Amplitude value + * @param outOrOutMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional, if outMax doesn't get passed in, it will be [min, max] with all channels combined instead). + * @param outMax The output maximum value from the analyzer, indices is in channels (Optional). + * @return Output of amplitude from given position. **/ public function analyze(startPos:Float, endPos:Float, ?outOrOutMin:Array, ?outMax:Array):Float { + var hasOut = outOrOutMin != null; + var hasTwoOut = hasOut && outMax != null; + + if (hasTwoOut) for (i in 0...buffer.channels) __min[i] = __max[i] = 0; __minByte = __maxByte = 0; - if (outOrOutMin != null) { + __check(); + __read(startPos, endPos, hasTwoOut ? __analyzeCallback : __analyzeCallbackSimple); + + if (hasOut) { var f:Float; - if (outMax != null) { - for (i in 0...buffer.channels) __min[i] = __max[i] = 0; - - read(startPos, endPos, __analyzeCallback); - - for (i in 0...buffer.channels) { - if (outOrOutMin[i] < (f = __min[i] / byteSize)) outOrOutMin[i] = f; - if (outMax[i] < (f = __max[i] / byteSize)) outMax[i] = f; - } + if (hasTwoOut) for (i in 0...buffer.channels) { + if (outOrOutMin[i] < (f = __min[i] / byteSize)) outOrOutMin[i] = f; + if (outMax[i] < (f = __max[i] / byteSize)) outMax[i] = f; } else { - read(startPos, endPos, __analyzeCallbackSimple); - outOrOutMin.resize(2); if (outOrOutMin[0] < (f = __minByte / byteSize)) outOrOutMin[0] = f; if (outOrOutMin[1] < (f = __maxByte / byteSize)) outOrOutMin[1] = f; } } - else - read(startPos, endPos, __analyzeCallbackSimple); return (__maxByte + __minByte) / byteSize; } @@ -121,9 +235,54 @@ class AudioAnalyzer { function __analyzeCallbackSimple(b:Int, c:Int):Void ((b > __maxByte) ? (__maxByte = b) : (if (-b > __minByte) (__minByte = -b))); + /** + * Gets samples from startPos with given length of samples. + * @param startPos Start Position of the FlxSound in milliseconds. + * @param length Length of Samples. + * @param mono Merge all of the byte channels of samples in one channel instead (Optional). + * @param Output that gets passed into this function (Optional). + * @return Output of + **/ + public function getSamples(startPos:Float, length:Int, mono = true, ?output:Array):Array { + ((output == null) ? (__sampleOutput = output = []) : (__sampleOutput = output)).resize((__sampleOutputLength = length) * (mono ? 1 : buffer.channels)); + __sampleIndex = 0; + + __check(); + __read(startPos, startPos + (length / __toBits), mono ? __getSamplesCallbackMerge : __getSamplesCallback); + + __sampleOutput = null; + return output; + } + + // TODO: check later if this formula is correct. + function __getSamplesCallbackMerge(b:Int, c:Int):Void if (__sampleIndex < __sampleOutputLength) { + if (c == 0) __sampleOutput[__sampleIndex] = b / buffer.channels / byteSize; + else if (c == buffer.channels) { + __sampleOutput[__sampleIndex] += b / buffer.channels / byteSize; + __sampleIndex++; + } + else + __sampleOutput[__sampleIndex] += b / buffer.channels / byteSize; + } + + function __getSamplesCallback(b:Int, c:Int):Void if (__sampleIndex < __sampleOutputLength) { + __sampleOutput[__sampleIndex] = b / byteSize; + __sampleIndex++; + } + + /** + * Read an attached FlxSound from startPos to endPos in milliseconds with a callback. + * @param startPos Start Position of the FlxSound in milliseconds. + * @param endPos End Position of the FlxSound in milliseconds. + * @param callback Int->Int->Void Byte->Channels->Void Callback to get the byte of a sample. + **/ public function read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { __check(); - if (buffer.data != null) __read(startPos, endPos, callback); + __read(startPos, endPos, callback); + } + + inline function __read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { + if (buffer.data != null) __readData(startPos, endPos, callback); #if lime_cffi else if (__canReadStream() && (startPos += __readStream(startPos, endPos, callback)) >= endPos) return; #if lime_vorbis @@ -132,7 +291,7 @@ class AudioAnalyzer { #end } - inline function __read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { + inline function __readData(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { var pos = Math.floor(startPos * __toBits), end = Math.floor(endPos * __toBits), c = 0; pos -= pos % __sampleSize; end -= end % __sampleSize; @@ -212,44 +371,4 @@ class AudioAnalyzer { } #end #end -} - -private typedef AudioAnalyzerCallback = Int->Int->Void; - -/* -abstract Complex(Array) { - public var real(get, never):Float; - inline function get_real() return this[0]; - - public var imag(get, never):Float; - inline function get_imag() return this[1]; - - public inline function new(real:Float, imag:Float) this = [real, imag]; - public static inline function fromReal(real:Float):Complex return new Complex(real, 0); - public static inline function exp(w:Float):Complex return new Complex(Math.cos(w), Math.sin(w)); - - public var angle(get, never):Float; - inline function get_angle() return Math.atan2(imag, real); - - public var magnitude(get, never):Float; - inline function get_magnitude():Float return Math.sqrt(real*real + imag*imag); - - @:op(A + B) - public inline function add(rhs:Complex):Complex return new Complex(real + rhs.real, imag + rhs.imag); - - @:op(A - B) - public inline function sub(rhs:Complex):Complex return new Complex(real - rhs.real, imag - rhs.imag); - - @:op(A * B) - public inline function mult(rhs:Complex):Complex - return new Complex(real*rhs.real - imag*rhs.imag, real*rhs.imag + imag*rhs.real); - - @:op(A / B) - public inline function div(rhs:Complex):Complex { - var m = rhs.magnitude; - return new Complex((real*rhs.real + imag*rhs.imag) / m, (imag*rhs.real - real*rhs.imag) / m); - } - - public inline function conj():Complex return new Complex(real, -imag); -} -*/ \ No newline at end of file +} \ No newline at end of file From 888f4135e927343b890776741e3012709c5112d3 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Fri, 1 Aug 2025 11:58:13 +0700 Subject: [PATCH 07/11] Fix size out of bound --- source/openfl/media/SoundChannel.hx | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/source/openfl/media/SoundChannel.hx b/source/openfl/media/SoundChannel.hx index 4d9323f5..e63d5909 100644 --- a/source/openfl/media/SoundChannel.hx +++ b/source/openfl/media/SoundChannel.hx @@ -168,6 +168,12 @@ import lime.media.openal.AL; size = backend.bufferSizes[i = backend.bufferSizes.length - backend.queuedBuffers]; buf = backend.bufferDatas[i].buffer; pos -= Math.floor(backend.bufferTimes[i] * buffer.sampleRate * buffer.channels * wordSize); + while (pos > size) { + if (++i >= backend.bufferSizes.length) return false; + pos -= size; + buf = backend.bufferDatas[i].buffer; + size = backend.bufferSizes[i]; + } } else #end { @@ -183,10 +189,10 @@ import lime.media.openal.AL; if (c % 2 == 0) ((b > leftMax) ? (leftMax = b) : (if ((b = -b) > leftMin) (leftMin = b))); else ((b > rightMax) ? (rightMax = b) : (if ((b = -b) > rightMin) (rightMin = b))); if ((pos += wordSize) >= size) #if lime_cffi { - if (!backend.streamed || ++i >= backend.bufferDatas.length) break; - size = backend.bufferSizes[i]; - buf = backend.bufferDatas[i].buffer; + if (!backend.streamed || ++i >= backend.bufferSizes.length) break; pos = 0; + buf = backend.bufferDatas[i].buffer; + size = backend.bufferSizes[i]; } #else break; #end From aabfb0f3cd8890e197978dd8f81535a6ad0b4c67 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Fri, 1 Aug 2025 11:58:21 +0700 Subject: [PATCH 08/11] it works? --- source/funkin/backend/utils/AudioAnalyzer.hx | 253 +++++++++++++------ 1 file changed, 181 insertions(+), 72 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index b05ad285..67206dd6 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -24,7 +24,7 @@ class AudioAnalyzer { if (wordSize == 2) return inline ArrayBufferIO.getInt16(buf, pos); else if (wordSize == 3) { var b = inline ArrayBufferIO.getUint16(buf, pos) | (buf.get(pos + 2) << 16); - if (b > 8388608) return b - 16777216; + if (b & 0x800000 != 0) return b - 0x1000000; else return b; } else if (wordSize == 4) return inline ArrayBufferIO.getInt32(buf, pos); @@ -33,7 +33,7 @@ class AudioAnalyzer { public var sound:FlxSound; public var buffer(default, null):AudioBuffer; - public var fftSamples(default, set):Int; + public var fftN(default, set):Int; public var byteSize(default, null):Int; @@ -59,11 +59,14 @@ class AudioAnalyzer { var __sampleOutput:Array; // fft + var __N2:Int; var __logN:Int; var __freqSamples:Array; var __reverseIndices:Array = []; - var __rootReals:Array = []; - var __rootImags:Array = []; + var __factors:Array = []; + var __windows:Array = []; + var __twiddleReals:Array = []; + var __twiddleImags:Array = []; var __freqReals:Array = []; var __freqImags:Array = []; @@ -74,9 +77,9 @@ class AudioAnalyzer { * Creates an analyzer for specified FlxSound * @param sound An FlxSound to analyze. **/ - public function new(sound:FlxSound, fftSamples = 512) { + public function new(sound:FlxSound, fftN = 2048) { this.sound = sound; - this.fftSamples = fftSamples; + this.fftN = fftN; __check(); } @@ -90,23 +93,50 @@ class AudioAnalyzer { __max.resize(buffer.channels); } - inline function set_fftSamples(v:Int):Int { - if (fftSamples == (fftSamples = v)) return v; - - var half = v >> 1, a, c; - __logN = Math.floor(Math.log(v) / Math.log(2)); - __reverseIndices.resize(half); - __rootReals.resize(half); - __rootImags.resize(half); - __freqReals.resize(half); - __freqImags.resize(half); - for (i in 0...half) { + inline function set_fftN(v:Int):Int { + if (fftN == (fftN = nextPow2(v))) return fftN; + + __logN = Math.floor(Math.log(fftN) / Math.log(2)); + __N2 = fftN >> 1; + __freqReals.resize(fftN); + __freqImags.resize(fftN); + __reverseIndices.resize(fftN); + __windows.resize(fftN); + __twiddleReals.resize(fftN); + __twiddleImags.resize(fftN); + + var f, a; + for (i in 0...fftN) { + f = i / (fftN - 1); + __windows[i] = 0.42 - 0.5 * Math.cos(Math.PI * 2 * f) + 0.08 * Math.cos(4 * Math.PI * f); __reverseIndices[i] = __bitReverse(i); - __rootReals[i] = Math.cos(a = 2 * Math.PI * i / v); - __rootImags[i] = Math.sin(a); + __twiddleReals[i] = Math.cos(a = 2 * Math.PI * i / fftN); + __twiddleImags[i] = Math.sin(a); } - return v; + __factors.resize(0); + + var inv = fftN; + while (inv > 0) { + if (inv % 4 == 0) { + __factors.push(4); + inv >>= 2; + } + else if (inv % 2 == 0) { + __factors.push(2); + inv >>= 1; + } + else + break; + } + + return fftN; + } + + inline function nextPow2(x:Int):Int { + var p = 1; + while (p < x) p <<= 1; + return p; } inline function __bitReverse(x:Int):Int { @@ -119,27 +149,39 @@ class AudioAnalyzer { return y; } - public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, minDb = -70.0, maxDb = -20.0):Array { + public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, dbRange = 50.0, minFreq = 20.0, maxFreq = 20000.0):Array { __frequencies = getFrequencies(startPos, __frequencies); if (levels == null) levels = []; levels.resize(barCount); - var x1 = 0.5, i1 = 1, n = fftSamples >> 1; - var x2, i2, v; - for (i in 0...barCount) { - i2 = Math.floor(x2 = Math.pow(n, (i + 1) / barCount) - 0.5); + var logMin = Math.log(minFreq), logMax = Math.log(maxFreq); + inline function calculateScale(i:Int) + return CoolUtil.bound(Math.exp(logMin + ((logMax - logMin) * i / barCount)) * fftN / buffer.sampleRate, 0, __N2 - 1); - if (i2 < i1) v = __frequencies[i2] * (x2 - x1); + var s1 = calculateScale(0), s2; + var i1 = Math.floor(s1), i2; + var v, range; + for (i in 0...barCount) { + i2 = Math.ceil(s2 = calculateScale(i)); + + if ((range = s2 - s1) < 1) { + if (i2 == i1) v = __frequencies[i1] * range; + else v = (s1 + (s2 - s1) * (s1 - i1)) * range; + } else { - v = __frequencies[i1 - 1] * (i1 - x1); - while (i1 < i2) v += __frequencies[i1++]; - if (i2 < n) v += __frequencies[i2] * (x2 - i2); + v = __frequencies[i1] * (Math.ceil(s1) - i1); + while (++i1 < i2) v += __frequencies[i1]; + v += __frequencies[i2] * (s2 - Math.floor(s2)); } - i1 = Math.ceil(x1 = x2); + i1 = Math.floor(s1 = s2); + + if (v > 0) v = 20 * Math.log(v) / 2.302585092994046;//log(10) + else v = -200; + + v = CoolUtil.bound(1 + v / dbRange, 0, 1); - v = 20 * Math.log(v * barCount / 12) / 2.302585092994046/*log(10)*/ / (maxDb - minDb); if (delta > 0 && levels[i] > v) levels[i] -= (levels[i] - v) * delta; else levels[i] = v; } @@ -149,48 +191,106 @@ class AudioAnalyzer { /** * Gets frequencies from an attached FlxSound from startPos with samples - * @return Output of frequencies ranging from 1 to fftSamples/2. + * @return Output of frequencies ranging from 1 to fftN/2. **/ public function getFrequencies(startPos:Float, ?frequencies:Array):Array { - __freqSamples = getSamples(startPos, fftSamples, true, __freqSamples); - - var inv = fftSamples >> 1; - for (i in 0...inv) { - __freqReals[__reverseIndices[i]] = __freqSamples[i]; - __freqImags[i] = 0; - } - - var h = 1, g = 0, b = 0, r = 0; - var tr, ti, i1, i2; - while (inv > 0) { - h <<= 1; - while (g < fftSamples) { - while (b < h) { - i2 = (i1 = g + b) + h; - tr = __rootReals[r] * __freqReals[i2] - __rootImags[r] * __freqImags[i2]; - ti = __rootReals[r] * __freqImags[i2] + __rootImags[r] * __freqReals[i2]; - - __rootReals[i2] = __freqReals[i1] - tr; - __rootImags[i2] = __freqImags[i1] - ti; - __freqReals[i1] += tr; - __freqImags[i1] += ti; - - b++; - r += inv; - } - b = r = 0; - g += h; - } - g = 0; - - inv >>= 1; - } + __freqSamples = getSamples(startPos, fftN, true, __freqSamples); if (frequencies == null) frequencies = []; - frequencies.resize(inv = fftSamples >> 1); - - var mag; - for (i in 0...inv) frequencies[i] = (mag = Math.sqrt(__freqReals[i] * __freqReals[i] + __freqImags[i] * __freqImags[i])) < 0.001 ? 0 : mag; + frequencies.resize(__N2); + + if (fftN == 1) frequencies[0] = __freqSamples[0]; + else { + for (i in 0...fftN) { + __freqReals[__reverseIndices[i]] = __freqSamples[i] * __windows[i]; + __freqImags[i] = 0; + } + + // https://github.com/FunkinCrew/grig.audio/commit/8567c4dad34cfeaf2ff23fe12c3796f5db80685e + inline function butterfly4PointOptimized(i0:Int, i1:Int, i2:Int, i3:Int, w1_idx:Int, w2_idx:Int, w3_idx:Int) { + // Load input values + var x0r = __freqReals[i0]; + var x0i = __freqImags[i0]; + + // Apply twiddle factors to x1, x2, x3 + // x1 = workingData[i1] * twiddle1 + var x1r_raw = __freqReals[i1]; + var x1i_raw = __freqImags[i1]; + var tw1r = __twiddleReals[w1_idx]; + var tw1i = __twiddleImags[w1_idx]; + var x1r = x1r_raw * tw1r - x1i_raw * tw1i; + var x1i = x1r_raw * tw1i + x1i_raw * tw1r; + + // x2 = workingData[i2] * twiddle2 + var x2r_raw = __freqReals[i2]; + var x2i_raw = __freqImags[i2]; + var tw2r = __twiddleReals[w2_idx]; + var tw2i = __twiddleImags[w2_idx]; + var x2r = x2r_raw * tw2r - x2i_raw * tw2i; + var x2i = x2r_raw * tw2i + x2i_raw * tw2r; + + // x3 = workingData[i3] * twiddle3 + var x3r_raw = __freqReals[i3]; + var x3i_raw = __freqImags[i3]; + var tw3r = __twiddleReals[w3_idx]; + var tw3i = __twiddleImags[w3_idx]; + var x3r = x3r_raw * tw3r - x3i_raw * tw3i; + var x3i = x3r_raw * tw3i + x3i_raw * tw3r; + + // Compute intermediate values for 4-point DFT + var t0r = x0r + x2r; // (x0 + x2).real + var t0i = x0i + x2i; // (x0 + x2).imag + var t1r = x0r - x2r; // (x0 - x2).real + var t1i = x0i - x2i; // (x0 - x2).imag + var t2r = x1r + x3r; // (x1 + x3).real + var t2i = x1i + x3i; // (x1 + x3).imag + var t3r = x1r - x3r; // (x1 - x3).real + var t3i = x1i - x3i; // (x1 - x3).imag + + // Apply j multiplication: j * (a + jb) = -b + ja + var jt3r = -t3i; // j * t3.real = -t3.imag + var jt3i = t3r; // j * t3.imag = t3.real + + // Final 4-point DFT butterfly outputs + __freqReals[i0] = t0r + t2r; // X[k] + __freqImags[i0] = t0i + t2i; + __freqReals[i1] = t1r - jt3r; // X[k + N/4] + __freqImags[i1] = t1i - jt3i; + __freqReals[i2] = t0r - t2r; // X[k + N/2] + __freqImags[i2] = t0i - t2i; + __freqReals[i3] = t1r + jt3r; // X[k + 3N/4] + __freqImags[i3] = t1i + jt3i; + } + + inline function butterfly2PointOptimized(i0:Int, i1:Int, w_idx:Int) { + var tempr = __freqReals[i1] * __twiddleReals[w_idx] - __freqImags[i1] * __twiddleImags[w_idx]; + var tempi = __freqReals[i1] * __twiddleImags[w_idx] + __freqImags[i1] * __twiddleReals[w_idx]; + __freqReals[i1] = __freqReals[i0] - tempr; + __freqImags[i1] = __freqImags[i0] - tempi; + __freqReals[i0] += tempr; + __freqImags[i0] += tempi; + } + + var size = 1, n, s, start, t; + for (radix in __factors) { + n = Math.floor(fftN / (size *= radix)); + s = size >> (radix >> 1); + if (radix == 4) for (i in 0...n) { + start = i * size; + for (k in 0...s) + butterfly4PointOptimized(t = start + k, t = (t + s), t = (t + s), t = (t + s), + (k * n) % fftN, (2 * k * n) % fftN, (3 * k * n) % fftN); + } + else for (i in 0...n) { + start = i * size; + for (k in 0...s) butterfly2PointOptimized(t = start + k, t = (t + s), (k * n) % fftN); + } + } + + var inv = 1.0 / fftN; + frequencies[0] = Math.sqrt(__freqReals[0] * __freqReals[0] + __freqImags[0] * __freqImags[0]) * inv; + for (i in 1...__N2) frequencies[i] = 2 * Math.sqrt(__freqReals[i] * __freqReals[i] + __freqImags[i] * __freqImags[i]) * inv; + } return frequencies; } @@ -248,7 +348,7 @@ class AudioAnalyzer { __sampleIndex = 0; __check(); - __read(startPos, startPos + (length / __toBits), mono ? __getSamplesCallbackMerge : __getSamplesCallback); + __read(startPos, startPos + (length / __toBits * buffer.channels), mono ? __getSamplesCallbackMerge : __getSamplesCallback); __sampleOutput = null; return output; @@ -315,16 +415,25 @@ class AudioAnalyzer { var n = Math.floor((endPos - startPos) * __toBits); if (startPos >= time && startPos < backend.bufferTimes[backend.bufferSizes.length - 1] * 1000) { var pos = Math.floor((startPos - time) * __toBits), buf = backend.bufferDatas[i].buffer, size = backend.bufferSizes[i], c = 0; + while (pos > size) { + if (++i >= backend.bufferSizes.length) { + n = 0; + break; + } + pos -= size; + buf = backend.bufferDatas[i].buffer; + size = backend.bufferSizes[i]; + } pos -= pos % __sampleSize; while (n > 0) { callback(getByte(buf, pos, __wordSize), c); if (++c > buffer.channels) c = 0; if ((pos += __wordSize) >= size) { - if (++i >= backend.bufferDatas.length) break; + if (++i >= backend.bufferSizes.length) break; + pos = 0; buf = backend.bufferDatas[i].buffer; size = backend.bufferSizes[i]; - pos = 0; } n -= __wordSize; } From a9495eed5e7da7e756dca3571024a97743fd7647 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Fri, 1 Aug 2025 15:15:03 +0700 Subject: [PATCH 09/11] Uhhuhhh --- source/funkin/backend/utils/AudioAnalyzer.hx | 201 +++++++++---------- 1 file changed, 98 insertions(+), 103 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 67206dd6..2eeedb6b 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -77,7 +77,7 @@ class AudioAnalyzer { * Creates an analyzer for specified FlxSound * @param sound An FlxSound to analyze. **/ - public function new(sound:FlxSound, fftN = 2048) { + public function new(sound:FlxSound, fftN = 512) { this.sound = sound; this.fftN = fftN; __check(); @@ -108,26 +108,23 @@ class AudioAnalyzer { var f, a; for (i in 0...fftN) { f = i / (fftN - 1); - __windows[i] = 0.42 - 0.5 * Math.cos(Math.PI * 2 * f) + 0.08 * Math.cos(4 * Math.PI * f); + __windows[i] = 0.42 - 0.5 * Math.cos(2 * Math.PI * f) + 0.08 * Math.cos(4 * Math.PI * f); __reverseIndices[i] = __bitReverse(i); - __twiddleReals[i] = Math.cos(a = 2 * Math.PI * i / fftN); + __twiddleReals[i] = Math.cos(a = -2 * Math.PI * i / fftN); __twiddleImags[i] = Math.sin(a); } __factors.resize(0); var inv = fftN; - while (inv > 0) { - if (inv % 4 == 0) { - __factors.push(4); - inv >>= 2; - } - else if (inv % 2 == 0) { - __factors.push(2); - inv >>= 1; - } - else - break; + while (inv % 4 == 0) { + __factors.push(4); + inv >>= 2; + } + + while (inv % 2 == 0) { + __factors.push(2); + inv >>= 1; } return fftN; @@ -149,40 +146,37 @@ class AudioAnalyzer { return y; } - public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, dbRange = 50.0, minFreq = 20.0, maxFreq = 20000.0):Array { + public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, minDb = -70, maxDb = -10, minFreq = 20.0, maxFreq = 22000.0):Array { __frequencies = getFrequencies(startPos, __frequencies); if (levels == null) levels = []; levels.resize(barCount); var logMin = Math.log(minFreq), logMax = Math.log(maxFreq); + var logRange = logMax - logMin, dbRange = maxDb - minDb; inline function calculateScale(i:Int) - return CoolUtil.bound(Math.exp(logMin + ((logMax - logMin) * i / barCount)) * fftN / buffer.sampleRate, 0, __N2 - 1); + return CoolUtil.bound(Math.exp(logMin + (logRange * i / (barCount + 1))) * fftN / buffer.sampleRate, 0, __N2 - 1); var s1 = calculateScale(0), s2; var i1 = Math.floor(s1), i2; var v, range; for (i in 0...barCount) { - i2 = Math.ceil(s2 = calculateScale(i)); - - if ((range = s2 - s1) < 1) { + if ((range = (s2 = calculateScale(i + 1)) - s1) < 1) { + i2 = Math.ceil(s2); if (i2 == i1) v = __frequencies[i1] * range; - else v = (s1 + (s2 - s1) * (s1 - i1)) * range; + else v = (__frequencies[i1] + (__frequencies[i2] - __frequencies[i1]) * (s1 - i1)) * range; } else { v = __frequencies[i1] * (Math.ceil(s1) - i1); - while (++i1 < i2) v += __frequencies[i1]; - v += __frequencies[i2] * (s2 - Math.floor(s2)); + if (i1 != (i2 = Math.floor(s2))) { + while (++i1 < i2) v += __frequencies[i1]; + v += __frequencies[i2] * (s2 - Math.floor(s2)); + } } - i1 = Math.floor(s1 = s2); - if (v > 0) v = 20 * Math.log(v) / 2.302585092994046;//log(10) - else v = -200; - - v = CoolUtil.bound(1 + v / dbRange, 0, 1); - - if (delta > 0 && levels[i] > v) levels[i] -= (levels[i] - v) * delta; + v = ((20 * Math.log(v) / 2.302585092994046) - minDb) / dbRange; + if (delta > 0 && delta < 1 && v < levels[i]) levels[i] -= Math.pow(levels[i] - v, 2) * delta; else levels[i] = v; } @@ -194,6 +188,71 @@ class AudioAnalyzer { * @return Output of frequencies ranging from 1 to fftN/2. **/ public function getFrequencies(startPos:Float, ?frequencies:Array):Array { + // https://github.com/FunkinCrew/grig.audio/commit/8567c4dad34cfeaf2ff23fe12c3796f5db80685e + inline function butterfly4PointOptimized(i0:Int, i1:Int, i2:Int, i3:Int, w1_idx:Int, w2_idx:Int, w3_idx:Int) { + // Load input values + var x0r = __freqReals[i0]; + var x0i = __freqImags[i0]; + + // Apply twiddle factors to x1, x2, x3 + // x1 = workingData[i1] * twiddle1 + var x1r_raw = __freqReals[i1]; + var x1i_raw = __freqImags[i1]; + var tw1r = __twiddleReals[w1_idx]; + var tw1i = __twiddleImags[w1_idx]; + var x1r = x1r_raw * tw1r - x1i_raw * tw1i; + var x1i = x1r_raw * tw1i + x1i_raw * tw1r; + + // x2 = workingData[i2] * twiddle2 + var x2r_raw = __freqReals[i2]; + var x2i_raw = __freqImags[i2]; + var tw2r = __twiddleReals[w2_idx]; + var tw2i = __twiddleImags[w2_idx]; + var x2r = x2r_raw * tw2r - x2i_raw * tw2i; + var x2i = x2r_raw * tw2i + x2i_raw * tw2r; + + // x3 = workingData[i3] * twiddle3 + var x3r_raw = __freqReals[i3]; + var x3i_raw = __freqImags[i3]; + var tw3r = __twiddleReals[w3_idx]; + var tw3i = __twiddleImags[w3_idx]; + var x3r = x3r_raw * tw3r - x3i_raw * tw3i; + var x3i = x3r_raw * tw3i + x3i_raw * tw3r; + + // Compute intermediate values for 4-point DFT + var t0r = x0r + x2r; // (x0 + x2).real + var t0i = x0i + x2i; // (x0 + x2).imag + var t1r = x0r - x2r; // (x0 - x2).real + var t1i = x0i - x2i; // (x0 - x2).imag + var t2r = x1r + x3r; // (x1 + x3).real + var t2i = x1i + x3i; // (x1 + x3).imag + var t3r = x1r - x3r; // (x1 - x3).real + var t3i = x1i - x3i; // (x1 - x3).imag + + // Apply j multiplication: j * (a + jb) = -b + ja + var jt3r = -t3i; // j * t3.real = -t3.imag + var jt3i = t3r; // j * t3.imag = t3.real + + // Final 4-point DFT butterfly outputs + __freqReals[i0] = t0r + t2r; // X[k] + __freqImags[i0] = t0i + t2i; + __freqReals[i1] = t1r - jt3r; // X[k + N/4] + __freqImags[i1] = t1i - jt3i; + __freqReals[i2] = t0r - t2r; // X[k + N/2] + __freqImags[i2] = t0i - t2i; + __freqReals[i3] = t1r + jt3r; // X[k + 3N/4] + __freqImags[i3] = t1i + jt3i; + } + + inline function butterfly2PointOptimized(i0:Int, i1:Int, w_idx:Int) { + var tempr = __freqReals[i1] * __twiddleReals[w_idx] - __freqImags[i1] * __twiddleImags[w_idx]; + var tempi = __freqReals[i1] * __twiddleImags[w_idx] + __freqImags[i1] * __twiddleReals[w_idx]; + __freqReals[i1] = __freqReals[i0] - tempr; + __freqImags[i1] = __freqImags[i0] - tempi; + __freqReals[i0] += tempr; + __freqImags[i0] += tempi; + } + __freqSamples = getSamples(startPos, fftN, true, __freqSamples); if (frequencies == null) frequencies = []; @@ -201,89 +260,26 @@ class AudioAnalyzer { if (fftN == 1) frequencies[0] = __freqSamples[0]; else { + var n; for (i in 0...fftN) { - __freqReals[__reverseIndices[i]] = __freqSamples[i] * __windows[i]; - __freqImags[i] = 0; + n = __reverseIndices[i]; + __freqReals[n] = __freqSamples[i] * __windows[i]; + __freqImags[n] = 0; } - // https://github.com/FunkinCrew/grig.audio/commit/8567c4dad34cfeaf2ff23fe12c3796f5db80685e - inline function butterfly4PointOptimized(i0:Int, i1:Int, i2:Int, i3:Int, w1_idx:Int, w2_idx:Int, w3_idx:Int) { - // Load input values - var x0r = __freqReals[i0]; - var x0i = __freqImags[i0]; - - // Apply twiddle factors to x1, x2, x3 - // x1 = workingData[i1] * twiddle1 - var x1r_raw = __freqReals[i1]; - var x1i_raw = __freqImags[i1]; - var tw1r = __twiddleReals[w1_idx]; - var tw1i = __twiddleImags[w1_idx]; - var x1r = x1r_raw * tw1r - x1i_raw * tw1i; - var x1i = x1r_raw * tw1i + x1i_raw * tw1r; - - // x2 = workingData[i2] * twiddle2 - var x2r_raw = __freqReals[i2]; - var x2i_raw = __freqImags[i2]; - var tw2r = __twiddleReals[w2_idx]; - var tw2i = __twiddleImags[w2_idx]; - var x2r = x2r_raw * tw2r - x2i_raw * tw2i; - var x2i = x2r_raw * tw2i + x2i_raw * tw2r; - - // x3 = workingData[i3] * twiddle3 - var x3r_raw = __freqReals[i3]; - var x3i_raw = __freqImags[i3]; - var tw3r = __twiddleReals[w3_idx]; - var tw3i = __twiddleImags[w3_idx]; - var x3r = x3r_raw * tw3r - x3i_raw * tw3i; - var x3i = x3r_raw * tw3i + x3i_raw * tw3r; - - // Compute intermediate values for 4-point DFT - var t0r = x0r + x2r; // (x0 + x2).real - var t0i = x0i + x2i; // (x0 + x2).imag - var t1r = x0r - x2r; // (x0 - x2).real - var t1i = x0i - x2i; // (x0 - x2).imag - var t2r = x1r + x3r; // (x1 + x3).real - var t2i = x1i + x3i; // (x1 + x3).imag - var t3r = x1r - x3r; // (x1 - x3).real - var t3i = x1i - x3i; // (x1 - x3).imag - - // Apply j multiplication: j * (a + jb) = -b + ja - var jt3r = -t3i; // j * t3.real = -t3.imag - var jt3i = t3r; // j * t3.imag = t3.real - - // Final 4-point DFT butterfly outputs - __freqReals[i0] = t0r + t2r; // X[k] - __freqImags[i0] = t0i + t2i; - __freqReals[i1] = t1r - jt3r; // X[k + N/4] - __freqImags[i1] = t1i - jt3i; - __freqReals[i2] = t0r - t2r; // X[k + N/2] - __freqImags[i2] = t0i - t2i; - __freqReals[i3] = t1r + jt3r; // X[k + 3N/4] - __freqImags[i3] = t1i + jt3i; - } - - inline function butterfly2PointOptimized(i0:Int, i1:Int, w_idx:Int) { - var tempr = __freqReals[i1] * __twiddleReals[w_idx] - __freqImags[i1] * __twiddleImags[w_idx]; - var tempi = __freqReals[i1] * __twiddleImags[w_idx] + __freqImags[i1] * __twiddleReals[w_idx]; - __freqReals[i1] = __freqReals[i0] - tempr; - __freqImags[i1] = __freqImags[i0] - tempi; - __freqReals[i0] += tempr; - __freqImags[i0] += tempi; - } - - var size = 1, n, s, start, t; + var size = 1, s2, start, t; for (radix in __factors) { n = Math.floor(fftN / (size *= radix)); - s = size >> (radix >> 1); + s2 = size >> (radix >> 1); if (radix == 4) for (i in 0...n) { start = i * size; - for (k in 0...s) - butterfly4PointOptimized(t = start + k, t = (t + s), t = (t + s), t = (t + s), + for (k in 0...s2) + butterfly4PointOptimized(t = start + k, t = (t + s2), t = (t + s2), t = (t + s2), (k * n) % fftN, (2 * k * n) % fftN, (3 * k * n) % fftN); } else for (i in 0...n) { start = i * size; - for (k in 0...s) butterfly2PointOptimized(t = start + k, t = (t + s), (k * n) % fftN); + for (k in 0...s2) butterfly2PointOptimized(t = start + k, t = (t + s2), (k * n) % fftN); } } @@ -344,7 +340,7 @@ class AudioAnalyzer { * @return Output of **/ public function getSamples(startPos:Float, length:Int, mono = true, ?output:Array):Array { - ((output == null) ? (__sampleOutput = output = []) : (__sampleOutput = output)).resize((__sampleOutputLength = length) * (mono ? 1 : buffer.channels)); + ((output == null) ? (__sampleOutput = output = []) : (__sampleOutput = output)).resize(__sampleOutputLength = length * (mono ? 1 : buffer.channels)); __sampleIndex = 0; __check(); @@ -354,7 +350,6 @@ class AudioAnalyzer { return output; } - // TODO: check later if this formula is correct. function __getSamplesCallbackMerge(b:Int, c:Int):Void if (__sampleIndex < __sampleOutputLength) { if (c == 0) __sampleOutput[__sampleIndex] = b / buffer.channels / byteSize; else if (c == buffer.channels) { From f3f517d0a8d4c11d2b104476d8c66a8c59df2f3a Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Fri, 1 Aug 2025 15:25:26 +0700 Subject: [PATCH 10/11] Disable 4 factoring, it looks ass with it --- source/funkin/backend/utils/AudioAnalyzer.hx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 2eeedb6b..42db432c 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -117,10 +117,10 @@ class AudioAnalyzer { __factors.resize(0); var inv = fftN; - while (inv % 4 == 0) { + /*while (inv % 4 == 0) { __factors.push(4); inv >>= 2; - } + }*/ while (inv % 2 == 0) { __factors.push(2); From e0a26fdbd2c32b60f35ffabafc2fdb34f0ce5fc8 Mon Sep 17 00:00:00 2001 From: Ralty <78720179+Raltyro@users.noreply.github.com> Date: Fri, 1 Aug 2025 15:45:39 +0700 Subject: [PATCH 11/11] Documentation for AudioAnalyzer --- source/funkin/backend/utils/AudioAnalyzer.hx | 82 +++++++++++++++----- 1 file changed, 61 insertions(+), 21 deletions(-) diff --git a/source/funkin/backend/utils/AudioAnalyzer.hx b/source/funkin/backend/utils/AudioAnalyzer.hx index 42db432c..3f552c55 100644 --- a/source/funkin/backend/utils/AudioAnalyzer.hx +++ b/source/funkin/backend/utils/AudioAnalyzer.hx @@ -18,23 +18,48 @@ typedef AudioAnalyzerCallback = Int->Int->Void; * * FlxSound.amplitude does work in CNE so if any case if your only checking for peak of current * time, use that instead. -**/ + */ class AudioAnalyzer { - public static function getByte(buf:ArrayBuffer, pos:Int, wordSize:Int):Int { - if (wordSize == 2) return inline ArrayBufferIO.getInt16(buf, pos); + /** + * Get bytes from an audio buffer with specified position and wordSize + * @param buffer The audio buffer to get byte from. + * @param position The specified position to get the byte from the audio buffer. + * @param wordSize How many bytes to get with to one byte (Usually it's bitsPerSample / 8 or bitsPerSample >> 3). + * @return Byte from the audio buffer with specified position. + */ + public static function getByte(buffer:ArrayBuffer, position:Int, wordSize:Int):Int { + if (wordSize == 2) return inline ArrayBufferIO.getInt16(buffer, position); else if (wordSize == 3) { - var b = inline ArrayBufferIO.getUint16(buf, pos) | (buf.get(pos + 2) << 16); + var b = inline ArrayBufferIO.getUint16(buffer, position) | (buffer.get(position + 2) << 16); if (b & 0x800000 != 0) return b - 0x1000000; else return b; } - else if (wordSize == 4) return inline ArrayBufferIO.getInt32(buf, pos); - else return inline ArrayBufferIO.getUint8(buf, pos) - 128; + else if (wordSize == 4) return inline ArrayBufferIO.getInt32(buffer, position); + else return inline ArrayBufferIO.getUint8(buffer, position) - 128; } + /** + * The current sound to analyze. + */ public var sound:FlxSound; - public var buffer(default, null):AudioBuffer; + + /** + * How much samples for the fft to get. + * Usually for getting the levels or frequencies of the sound. + * + * Has to be power of two, or it won't work. + */ public var fftN(default, set):Int; + /** + * The current buffer from sound. + */ + public var buffer(default, null):AudioBuffer; + + /** + * The current byteSize from buffer. + * Example the byteSize of 16 BitsPerSample is 32768 (1 << 16-1) + */ public var byteSize(default, null):Int; var __toBits:Float; @@ -76,8 +101,9 @@ class AudioAnalyzer { /** * Creates an analyzer for specified FlxSound * @param sound An FlxSound to analyze. - **/ - public function new(sound:FlxSound, fftN = 512) { + * @param fftN How much samples for fft to get (Optional, default 2048). + */ + public function new(sound:FlxSound, fftN = 2048) { this.sound = sound; this.fftN = fftN; __check(); @@ -146,7 +172,19 @@ class AudioAnalyzer { return y; } - public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, minDb = -70, maxDb = -10, minFreq = 20.0, maxFreq = 22000.0):Array { + /** + * Gets levels from an attached FlxSound from startPos, basically a minimized of frequencies. + * @param startPos Start Position to get from sound in milliseconds. + * @param barCount How much bars to get. + * @param levels The output for getting the values, to avoid memory leaks (Optional). + * @param delta How much delta for smoothen the values from the previous levels values (Optional). + * @param minDb The minimum decibels to cap (Optional, default -70.0). + * @param maxDb The maximum decibels to cap (Optional, default -10.0). + * @param minFreq The minimum frequency to cap (Optional, default 20.0). + * @param maxFreq The maximum frequency to cap (Optional, default 22000.0). + * @return Output of levels/bars + */ + public function getLevels(startPos:Float, barCount:Int, ?levels:Array, delta = 0.0, minDb = -70.0, maxDb = -10.0, minFreq = 20.0, maxFreq = 22000.0):Array { __frequencies = getFrequencies(startPos, __frequencies); if (levels == null) levels = []; @@ -184,9 +222,11 @@ class AudioAnalyzer { } /** - * Gets frequencies from an attached FlxSound from startPos with samples - * @return Output of frequencies ranging from 1 to fftN/2. - **/ + * Gets frequencies from an attached FlxSound from startPos. + * @param startPos Start Position to get from sound in milliseconds. + * @param frequencies The output for getting the frequencies, to avoid memory leaks (Optional). + * @return Output of frequencies + */ public function getFrequencies(startPos:Float, ?frequencies:Array):Array { // https://github.com/FunkinCrew/grig.audio/commit/8567c4dad34cfeaf2ff23fe12c3796f5db80685e inline function butterfly4PointOptimized(i0:Int, i1:Int, i2:Int, i3:Int, w1_idx:Int, w2_idx:Int, w3_idx:Int) { @@ -293,12 +333,12 @@ class AudioAnalyzer { /** * Analyzes an attached FlxSound from startPos to endPos in milliseconds to get the amplitudes. - * @param startPos Start Position of the FlxSound in milliseconds. - * @param endPos End Position of the FlxSound in milliseconds. + * @param startPos Start Position to get from sound in milliseconds. + * @param endPos End Position to get from sound in milliseconds. * @param outOrOutMin The output minimum value from the analyzer, indices is in channels (0 to -0.5 -> 0 to 0.5) (Optional, if outMax doesn't get passed in, it will be [min, max] with all channels combined instead). * @param outMax The output maximum value from the analyzer, indices is in channels (Optional). * @return Output of amplitude from given position. - **/ + */ public function analyze(startPos:Float, endPos:Float, ?outOrOutMin:Array, ?outMax:Array):Float { var hasOut = outOrOutMin != null; var hasTwoOut = hasOut && outMax != null; @@ -333,12 +373,12 @@ class AudioAnalyzer { /** * Gets samples from startPos with given length of samples. - * @param startPos Start Position of the FlxSound in milliseconds. + * @param startPos Start Position to get from sound in milliseconds. * @param length Length of Samples. * @param mono Merge all of the byte channels of samples in one channel instead (Optional). * @param Output that gets passed into this function (Optional). * @return Output of - **/ + */ public function getSamples(startPos:Float, length:Int, mono = true, ?output:Array):Array { ((output == null) ? (__sampleOutput = output = []) : (__sampleOutput = output)).resize(__sampleOutputLength = length * (mono ? 1 : buffer.channels)); __sampleIndex = 0; @@ -367,10 +407,10 @@ class AudioAnalyzer { /** * Read an attached FlxSound from startPos to endPos in milliseconds with a callback. - * @param startPos Start Position of the FlxSound in milliseconds. - * @param endPos End Position of the FlxSound in milliseconds. + * @param startPos Start Position to get from sound in milliseconds. + * @param endPos End Position to get from sound in milliseconds. * @param callback Int->Int->Void Byte->Channels->Void Callback to get the byte of a sample. - **/ + */ public function read(startPos:Float, endPos:Float, callback:AudioAnalyzerCallback) { __check(); __read(startPos, endPos, callback);