--- title: "AudioPitchShift" language: "en" type: "Symbol" summary: "AudioPitchShift[audio, r] applies pitch shifting to audio by the ratio r, shifting every frequency f to r f. AudioPitchShift[video, r] applies pitch shifting to the first audio track in video." keywords: - Pitch Shift - Phase Vocoder - Pitch - Frequency Shift - audio pitch shift - pitch shifting - time stretch - change audio pitch canonical_url: "https://reference.wolfram.com/language/ref/AudioPitchShift.html" source: "Wolfram Language Documentation" related_guides: - title: "Audio Editing" link: "https://reference.wolfram.com/language/guide/AudioEditing.en.md" - title: "Signal Processing" link: "https://reference.wolfram.com/language/guide/SignalProcessing.en.md" - title: "Speech Computation" link: "https://reference.wolfram.com/language/guide/SpeechComputation.en.md" - title: "Sound and Sonification" link: "https://reference.wolfram.com/language/guide/SoundAndSonification.en.md" - title: "Video Editing" link: "https://reference.wolfram.com/language/guide/VideoEditing.en.md" - title: "Video Computation: Update History" link: "https://reference.wolfram.com/language/guide/VideoComputation-UpdateHistory.en.md" - title: "Audio Processing" link: "https://reference.wolfram.com/language/guide/AudioProcessing.en.md" related_functions: - title: "AudioFrequencyShift" link: "https://reference.wolfram.com/language/ref/AudioFrequencyShift.en.md" - title: "AudioTimeStretch" link: "https://reference.wolfram.com/language/ref/AudioTimeStretch.en.md" - title: "PitchRecognize" link: "https://reference.wolfram.com/language/ref/PitchRecognize.en.md" - title: "AudioResample" link: "https://reference.wolfram.com/language/ref/AudioResample.en.md" - title: "AudioTrackApply" link: "https://reference.wolfram.com/language/ref/AudioTrackApply.en.md" --- # AudioPitchShift AudioPitchShift[audio, r] applies pitch shifting to audio by the ratio r, shifting every frequency $f$ to $r f$. AudioPitchShift[video, r] applies pitch shifting to the first audio track in video. ## Details and Options * ``AudioPitchShift`` shifts the pitch of the input audio signal without affecting the duration. [image] * The frequency ratio ``r`` can be given as a pitch interval, including: | | | | -------------------------- | --------------------------------------------------------------------------- | | Quantity[oct, "octaves"] | correspond to $r=2^{\text{\textit{oct}}}$ | | Quantity[sem, "semitones"] | correspond to $r=2^{\text{\textit{sem}}/12}$ | * The following options are supported: | | | | | --------------------- | --------- | ------------------------------------------------------------------------------------- | | Method | Automatic | the method to use | | PartitionGranularity | Automatic | control the partitioning used for the computation of the short-time Fourier transform | * By default, a phase vocoder algorithm is used. Use ``Method -> "Speech"`` for pitch shifting speech signals. In this case, while the fundamental frequency is changed, the formants of the signal are not modified. --- ## Examples (15) ### Basic Examples (2) Shift the audio pitch by a factor of two: ```wl In[1]:= AudioPitchShift[\!\(\*AudioBox["![Embedded Audio Player](audio://content-1il6e)"]\), 2] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-x63r6)"]\) ``` --- Shift the audio pitch by a semitone: ```wl In[1]:= AudioPitchShift[\!\(\*AudioBox["![Embedded Audio Player](audio://content-1il6e)"]\), Quantity[1, IndependentUnit["semitones"]]] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-r326p)"]\) ``` ### Scope (4) A shift by ``n`` octaves corresponds to a frequency ratio of ``2^n`` : ```wl In[1]:= a = Import["ExampleData/rule30.wav"]; In[2]:= AudioPitchShift[a, Quantity[1.5, IndependentUnit["octaves"]]] === AudioPitchShift[a, 2^1.5] Out[2]= True ``` --- A shift by ``s`` semitones corresponds to a frequency ratio of ``2^s / 12`` : ```wl In[1]:= a = Import["ExampleData/rule30.wav"]; In[2]:= AudioPitchShift[a, Quantity[10, IndependentUnit["semitones"]]] === AudioPitchShift[a, 2^10 / 12] Out[2]= True ``` --- ``AudioPitchShift`` can be used to change the pitch of a harmonically complex sound without altering its timbre and timing: ```wl In[1]:= a = Mean[Table[AudioGenerator[{"Sin", 200 * i}, 2], {i, 1, 6}]] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-eic2c)"]\) In[2]:= Periodogram[a, ScalingFunctions -> "Absolute", PlotRange -> {{0, 3000}, All}] Out[2]= [image] ``` The frequency ratio between the components of the audio signal is not changed: ```wl In[3]:= AudioPitchShift[a, Quantity[12, IndependentUnit["semitones"]]] Out[3]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-xasji)"]\) In[4]:= Periodogram[%, ScalingFunctions -> "Absolute", PlotRange -> {{0, 3000}, All}] Out[4]= [image] ``` --- Process the audio track of a video: ```wl In[1]:= AudioPitchShift[\!\(\*VideoBox["![Video Player: ExampleData/fish.mp4](video://content-2sfji)"]\), 2] Out[1]= \!\(\*VideoBox[...]\) ``` ### Options (4) #### Method (1) Use ``Method -> "Speech"`` for speech signals to shift the pitch while maintaining other voice characteristics: ```wl In[1]:= a = \!\(\*AudioBox["![Embedded Audio Player](audio://content-2pcu3)"]\); AudioPitchShift[a, 1.8, Method -> "Speech"] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-vv029)"]\) ``` Compare with the normal pitch shifting, which dramatically affects the timbre of the voice: ```wl In[2]:= AudioPitchShift[a, 1.8] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-578kf)"]\) ``` #### PartitionGranularity (3) Using a bigger partition size will improve the frequency response but will smear the transient components: ```wl In[1]:= a = Import["ExampleData/rule30.wav"]; AudioPitchShift[a, 1.1, PartitionGranularity -> Quantity[46, "Milliseconds"]] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-exgx5)"]\) In[2]:= AudioPitchShift[a, 1.1, PartitionGranularity -> Quantity[150, "Milliseconds"]] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-03whf)"]\) ``` --- Using a small offset value will in general increase the quality of the result at the cost of more computation time: ```wl In[1]:= a = Import["ExampleData/rule30.wav"]; AudioPitchShift[a, 1.1, PartitionGranularity -> {Quantity[46, "Milliseconds"], Quantity[20, "Milliseconds"]}] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-fu1wc)"]\) In[2]:= AudioPitchShift[a, 1.1, PartitionGranularity -> {Quantity[46, "Milliseconds"], Quantity[5, "Milliseconds"]}] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-0vxuv)"]\) ``` If only the partition size is specified, the offset is calculated as 25% of the partition size. --- Use a different window shape to change the quality of the result: ```wl In[1]:= a = Import["ExampleData/rule30.wav"]; AudioPitchShift[a, 1.1, PartitionGranularity -> {Quantity[46, "Milliseconds"], Quantity[10, "Milliseconds"], DirichletWindow}] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-925us)"]\) In[2]:= AudioPitchShift[a, 1.1, PartitionGranularity -> {Quantity[46, "Milliseconds"], Quantity[10, "Milliseconds"], BlackmanWindow}] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-925us)"]\) ``` By default, ``HannWindow`` is used. ### Applications (1) Shift the pitch of an audio object to match a target frequency: ```wl In[1]:= a = \!\(\*AudioBox["![Embedded Audio Player](audio://content-oxizp)"]\); ``` Extract the mean frequency with ``AudioLocalMeasurements`` : ```wl In[2]:= meanFreq = DeleteMissing@TimeSeries[AudioLocalMeasurements[a, "FundamentalFrequency"]]["Values"]//Mean Out[2]= 262.63 ``` Shift the audio to reach the target frequency: ```wl In[3]:= targetFreq = 1000; In[4]:= AudioPitchShift[a, targetFreq / meanFreq] Out[4]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-izyh6)"]\) In[5]:= Periodogram[{a, %}, 4000, PlotRange -> {{0, 4000}, All}] Out[5]= [image] ``` ### Properties & Relations (2) ``AudioPitchShift`` and ``AudioFrequencyShift`` modify the spectrum of the signal in different ways: ```wl In[1]:= a = Mean[{AudioGenerator[{"Sin", 200}, SampleRate -> 16000], AudioGenerator[{"Sin", 400}, SampleRate -> 16000]}] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-ggi7p)"]\) ``` ``AudioPitchShift`` shifts each frequency ``f`` in the signal by an amount proportional to the frequency itself (``f^' = f×shift``), thus preserving the timbre in the shifted signal: ```wl In[2]:= ps = AudioPitchShift[a, 4] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-9gtjf)"]\) ``` ``AudioFrequencyShift`` shifts every frequency ``f`` by the same amount (``f^' = f + shift``), changing dramatically the timbre in the shifted signal: ```wl In[3]:= fs = AudioFrequencyShift[a, 800] Out[3]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-zci51)"]\) In[4]:= Periodogram[{a, ps, fs}, PlotRange -> {{0, 2000}, All}, ScalingFunctions -> "Absolute", PlotLegends -> {"Original", "Pitch Shift", "Frequency Shift"}] Out[4]= [image] ``` --- Change the pitch of a recording by using a different sample rate: ```wl In[1]:= a = \!\(\*AudioBox["![Embedded Audio Player](audio://content-ctqtq)"]\); In[2]:= Audio[a, SampleRate -> 44100] Out[2]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-gqwq2)"]\) ``` ``AudioPitchShift`` changes the pitch without altering the timing: ```wl In[3]:= AudioPitchShift[a, 44100 / 22050] Out[3]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-k17zo)"]\) ``` ### Possible Issues (2) The quality of the result degrades with large shift parameters: ```wl In[1]:= AudioPitchShift[\!\(\*AudioBox["![Embedded Audio Player](audio://content-1il6e)"]\), 5.6] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-ubfox)"]\) ``` --- Using ``Method -> "Speech"`` for non-speech signals may not sound as expected: ```wl In[1]:= AudioPitchShift[\!\(\*AudioBox["![Embedded Audio Player](audio://content-3rb3d)"]\), 1.8, Method -> "Speech"] Out[1]= \!\(\*AudioBox["![Embedded Audio Player](audio://content-re1cz)"]\) ``` ## See Also * [`AudioFrequencyShift`](https://reference.wolfram.com/language/ref/AudioFrequencyShift.en.md) * [`AudioTimeStretch`](https://reference.wolfram.com/language/ref/AudioTimeStretch.en.md) * [`PitchRecognize`](https://reference.wolfram.com/language/ref/PitchRecognize.en.md) * [`AudioResample`](https://reference.wolfram.com/language/ref/AudioResample.en.md) * [`AudioTrackApply`](https://reference.wolfram.com/language/ref/AudioTrackApply.en.md) ## Related Guides * [Audio Editing](https://reference.wolfram.com/language/guide/AudioEditing.en.md) * [Signal Processing](https://reference.wolfram.com/language/guide/SignalProcessing.en.md) * [Speech Computation](https://reference.wolfram.com/language/guide/SpeechComputation.en.md) * [Sound and Sonification](https://reference.wolfram.com/language/guide/SoundAndSonification.en.md) * [Video Editing](https://reference.wolfram.com/language/guide/VideoEditing.en.md) * [Video Computation: Update History](https://reference.wolfram.com/language/guide/VideoComputation-UpdateHistory.en.md) * [Audio Processing](https://reference.wolfram.com/language/guide/AudioProcessing.en.md) ## History * [Introduced in 2016 (11.0)](https://reference.wolfram.com/language/guide/SummaryOfNewFeaturesIn110.en.md) \| [Updated in 2020 (12.1)](https://reference.wolfram.com/language/guide/SummaryOfNewFeaturesIn121.en.md) ▪ [2024 (14.1)](https://reference.wolfram.com/language/guide/SummaryOfNewFeaturesIn141.en.md)