{"id":94,"date":"2024-02-17T12:04:58","date_gmt":"2024-02-17T20:04:58","guid":{"rendered":"https:\/\/www.kerneltrick.com\/?p=94"},"modified":"2024-02-17T12:09:51","modified_gmt":"2024-02-17T20:09:51","slug":"time-stretch-audio-and-preserving-pitch","status":"publish","type":"post","link":"https:\/\/www.kerneltrick.com\/?p=94","title":{"rendered":"Time Stretch audio and preserving pitch"},"content":{"rendered":"

Here is a example program that will take an audio sample, stretch it out in time but will preserve the pitch. It’s based loosely on paulstretch.
Written in BlitzMax NG.\u00a0 The program is written for simplicity not speed.<\/p>\r\n

The important variables are:<\/p>\r\n

ScaleFactor : how much to stretch the audio, a value of 3 would be 3 times the original length.
WindowLen: Larger makes a cleaner sound but is slower to run. Must be a power of 2.
Number of threads to run: This should less that the number of logical cores on your cpu.<\/p>\r\n

Hit escape key to exit the program. Download windows exe, sourcecode and included audio file HERE<\/a> \"\"<\/a><\/p>\r\n\r\n

\r\n'\/\/ Time stretch audio roughly based on paulstretch by A.Woodard [aaron at kerneltrick.com]\r\n\r\nSuperStrict\r\nImport brl.glmax2d\r\nImport brl.threadpool\r\nImport pub.freeaudio\r\nImport brl.audio\r\nImport brl.audiosample\r\nImport brl.wavloader\r\nImport brl.oggloader\r\n\r\nAppTitle = \"Timestretch audio in BlitzMax\"\r\n\r\nType TTask Extends TRunnable\r\n\tGlobal mutex:TMutex = TMutex.Create()\r\n\tField Readpos:Int, Writepos:Int\r\n\tMethod New(Rp:Int, Wp:Int)\r\n\t\tReadPos = Rp\r\n\t\tWritepos = Wp\r\n\t\tPrint(\"task::new\")\r\n\tEnd Method\r\n\t\r\n\tMethod run()\r\n\t\tDoChunk(ReadPos, writepos, mutex)\r\n\t\tDelay 2\r\n\tEnd Method\r\nEnd Type\r\n\r\nSeedRnd(MilliSecs())\r\n\r\nConst TWOPI:Double = Pi * 2\r\nConst GW:Int = 800\r\nConst GH:Int = 600\r\nConst OrigAudioFile:String = \"dhm.wav\"\r\n\r\nGlobal Tpool:TThreadPoolExecutor = TThreadPoolExecutor.newFixedThreadPool(4)  '\/\/ Num threads should be less than cpu cores\r\nGlobal Origdata:Float[]\r\nGlobal ScaleFactor:Float = 4\t\t\t\t\t\t\t\t\t\t\t\t\t'\/\/ How much to stretch the audio\r\nGlobal Newdata:Float[(Origdata.length * scalefactor) + 1]\r\nGlobal WindowLen:Float = 1024 * 4\t\t\t\t\t\t\t\t\t\t\t\t'\/\/ Window size, larger sounds better but is slower\r\nGlobal Offset:Float = (WindowLen \/ 2) \/ (ScaleFactor)\r\nGlobal ReadPos:Int = 0\r\nGlobal N:Int = windowlen\r\nGlobal orig:TAudioSample\r\n\r\nGraphics GW, GH,0,10\r\n\r\nDoStretch()\r\n\r\nWhile Not KeyHit(KEY_ESCAPE)\r\n\tCls\r\n\t\tDrawData()\r\n\t\t'DrawText(Tpool.threadCount + \" \" + Tpool.threadsWorking, 10, 10)\r\n\t\t\r\n\t\tIf KeyHit(KEY_1) And Tpool.threadsWorking = 0 Then\r\n\t\t\tPlaySound(LoadSound(orig))\r\n\t\tEndIf\r\n\t\t\r\n\t\tIf KeyHit(KEY_2) And Tpool.threadsWorking = 0 Then\r\n\t\t\tLocal newdat:Short[Newdata.length]\r\n\t\t\tFor Local i:Int = 0 Until newdat.Length\r\n\t\t\t\tnewdat[i] = Newdata[i] * 32767\r\n\t\t\tNext\r\n\t\t\tLocal newsamp:TAudioSample = New TAudioSample()\r\n\t\t\tnewsamp.format = SF_MONO16LE\r\n\t\t\tnewsamp = newsamp.CreateStatic(Varptr newdat[0], newdat.Length, 44100, SF_MONO16LE)\r\n\t\t\tPlaySound(LoadSound(newsamp))\r\n\t\tEnd If\r\n\t\t\r\n\t\tIf Tpool.threadsWorking = 0 Then\r\n\t\t\tDrawText(\"Press 1 to play original sound, Press 2 to play stretched sound\", 10, 10)\r\n\t\tElse\r\n\t\t\tDrawText(\"Working, wait....\", 10, 10)\r\n\t\tEndIf\r\n\t\t\r\n\tFlip\r\nWend\r\n\r\n\r\n'-------------------------------------------------------------------------------------------------\r\nFunction DoStretch()\r\n\tLocal cPos:Int\r\n\tLocal count:Int = 0\r\n\tIf Not(Not (WindowLen & (WindowLen - 1))) Then Notify(\"windowlen must be pow\/2\", 1) ;End\r\n\torig:TAudioSample = LoadAudioSample(OrigAudioFile)\t'\/\/ Assumes Mono. Using ogg sample beacuse this func hates .wav\r\n\tIf Not orig Then Notify(\"Cant load audio\", 1) ;End\r\n\torig = orig.Convert(SF_MONO16LE)\r\n\tOrigdata = New Float[orig.Length]\r\n\tLocal p:Short Ptr = Short Ptr(orig.samples)\r\n\t\r\n\tFor Local i:Int = 0 Until orig.Length\r\n\t\tOrigdata[i] = 0.8 * ((Float(Short(32767 - p[0])) \/ 32767.0) - 1.0) \r\n\t\tp:+1\r\n\tNext\r\n\t\r\n\tNewdata = New Float[(Origdata.Length * ScaleFactor) + 1]\r\n\t\r\n\tWhile ReadPos + WindowLen < Origdata.Length\r\n\t\tCPos = ((N \/ 2) * Count)\r\n\t\t\r\n\t\tTpool.execute(New TTask(ReadPos, CPos))\r\n\t\t\r\n\t\tCount:+1\r\n\t\tReadPos:+Offset\r\n\tWend\r\nEnd Function\r\n\r\n\r\n'-------------------------------------------------------------------------------------------------\r\nFunction DoChunk(Readpos:Int, Writepos:Int, mut:TMutex)\r\n\tPrint \"HERE  \" + ReadPos + \" \" + Writepos\r\n\tLocal cp:Float[N]\r\n\tLocal sp:Float[N]\r\n\tLocal SpecBuff:Float[]\r\n\tLocal obuff:Float[]\r\n\tLocal isr:Float = 0.853\r\n\t\r\n\tSpecBuff = Origdata[Readpos..(Readpos + N)]\r\n\t\r\n\tFor Local i:Int = 0 Until N\r\n\t\tSpecBuff[i]:*Sine(Pi * i \/ (N))\r\n\tNext\r\n\tSpeccy(SpecBuff, -1, cp, sp)\r\n\tcp[0] = 0\r\n\tsp[0] = 0\r\n\tFor Local i:Int = 0 Until N \/ 2\r\n\t\tcp[i] = Sqr(cp[i] * cp[i] + sp[i] * sp[i])\r\n\t\tLocal p:Float = Rnd(-Pi, Pi)\r\n\t\tsp[i] = cp[i] * Sine(p)\r\n\t\tcp[i] = cp[i] * Cosine(p)\r\n\tNext\r\n\t\r\n\tSpeccy(SpecBuff, 1, cp, sp)\r\n\t\r\n\tFor Local i:Int = 0 Until N\r\n\t\tSpecBuff[i]:*Sine(Pi * i \/ N)\r\n\tNext\r\n\t\r\n\tmut.Lock()\t\r\n\t\tFor Local i:Int = 0 Until N '\/ 2\r\n\t\t\tNewdata[i + Writepos]:+(SpecBuff[i])\t\t\r\n\t\t\tNewdata[i + Writepos]:*(isr - (1.0 - isr) * Sine(Pi * i \/ N))\r\n\t\tNext\r\n\t\t\r\n\t\tFor Local i:Int = 0 Until N\r\n\t\t\tNewdata[i + Writepos] = ClampOne(Newdata[i + Writepos])\r\n\t\tNext\r\n\tmut.Unlock\r\n\t\r\n\tFunction Speccy(buff:Float[], s:Int = -1, cp:Float[], sp:Float[])\t'\/\/ Insanely slow !!!\r\n\t\tIf s = 1 Then\r\n\t\t\tMemClear(Varptr buff[0], SizeOf(1:Float) * N)\t\r\n\t\t\tFor Local k:Int = 0 Until N\r\n\t\t\t\tFor Local b:Int = 0 Until N \/ 2\r\n\t\t\t\t\tLocal a:Float = 2.0 * b * Pi * k \/ N\r\n\t\t\t\t\tbuff[k]:+(cp[b] * Cosine(a)) + ((-sp[b]) * Sine(a))\r\n\t\t\t\tNext\r\n\t\t\t\tbuff[k] = 2 * buff[k] \/ N\r\n\t\t\tNext\r\n\t\tElse\r\n\t\t\tMemClear(Varptr cp[0], SizeOf(1:Float) * N)\r\n\t\t\tMemClear(Varptr sp[0], SizeOf(1:Float) * N)\r\n\t\t\tFor Local k:Int = 0 Until N\t\t\r\n\t\t\t\tFor Local b:Int = 0 Until N \/ 2\r\n\t\t\t\t\tLocal a# = 2.0 * b * Pi * k \/ N\r\n\t\t\t\t\tsp[b]:+Buff[k] * s * Sine(a)\r\n\t\t\t\t\tcp[b]:+buff[k] * Cosine(a)\t\t\r\n\t\t\t\tNext\r\n\t\t\tNext\r\n\t\tEnd If\r\n\tEndFunction\r\nEnd Function\r\n'-------------------------------------------------------------------------------------------------\r\nFunction DrawData()\r\n\tLocal x:Float\r\n\tLocal xs:Float = GW \/ Float(Newdata.length)\r\n\tFor Local i:Int = 0 Until Origdata.Length\r\n\t\tDrawRect(x, 200 + Origdata[i] * 100, 2, 1)\r\n\t\tx:+xs\r\n\tNext\r\n\tx = 0\r\n\tFor Local i:Int = 0 Until Newdata.Length\r\n\t\tDrawRect(x, 400 + Newdata[i] * 100, 2, 1)\r\n\t\tx:+xs\r\n\tNext\r\n\tSetColor 255, 0, 255\r\n\tDrawText(\"Original\", 50, 200 - 5)\r\n\tDrawText(\"Streched\", 50, 400 - 5)\r\n\tSetColor 255, 255, 255\r\n\t'DrawRect(0, 200, 1000, 1)\r\n\tDelay 10\r\nEnd Function\r\n\r\n'-------------------------------------------------------------------------------------------------\r\nFunction Sine:Double(x:Double) inline\r\n    Local k:Int;Local y:Double;Local z:Double\r\n    z = x\r\n    z:*0.3183098861837907\r\n    z:+6755399441055744.0\r\n    k = Int Ptr(Varptr(z))[0]\r\n    z = k\r\n    z:*3.1415926535897932\r\n    x:-z;y = x;y:*x\r\n    z = 0.0073524681968701\r\n    z:*y\r\n    z:-0.1652891139701474\r\n    z:*y\r\n    z:+0.9996919862959676\r\n    x:*z;k:&1;k:+k;z = k;z:*x;x:-z\r\n    Return x\r\nEnd Function\r\n\r\n'-------------------------------------------------------------------------------------------------\r\nFunction Cosine:Double(x:Double) Inline\r\n\tReturn Sine((3.1415926535897932 \/ 2.0) - x)\r\nEnd Function\r\n'-------------------------------------------------------------------------------------------------\r\nFunction ClampOne:Double(x:Double) Inline\r\n\tReturn 0.5 * (Abs(x + 1) - Abs(x - 1))\r\nEnd Function\r\n'-------------------------------------------------------------------------------------------------\r\n\r\n\r\n\r\n<\/code><\/pre>\r\n","protected":false},"excerpt":{"rendered":"
Here is a example program that will take an audio sample, stretch it out in time but will preserve the pitch. It’s based loosely on paulstretch. Written in BlitzMax NG.\u00a0 The program is written for simplicity not speed. The important variables are: ScaleFactor : how much to stretch the audio, a value of 3 would… <\/p>\n
Read More<\/i><\/a><\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[21,6,1],"tags":[],"class_list":["post-94","post","type-post","status-publish","format-standard","hentry","category-audio","category-programming","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=\/wp\/v2\/posts\/94","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=94"}],"version-history":[{"count":0,"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=\/wp\/v2\/posts\/94\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=94"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=94"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kerneltrick.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=94"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}