SPTK

We are pleased to announce the release of SPTK 4.4!🚀 In this update, the raw2wav and wav2raw commands are back. We are also planning to add a resampling command in the near future. github.com/sp-nitech/SPTK #SpeechProcessing

Also, diffsptk version 3.4 is already out :) It includes WORLD analysis-synthesis and trainable STFT. If you're working with PyTorch, please give it a try. github.com/sp-nitech/diff… #PyTorch

We are pleased to announce the release of SPTK 4.4!🚀 In this update, the raw2wav and wav2raw commands are back. We are also planning to add a resampling command in the near future. github.com/sp-nitech/SPTK #SpeechProcessing

diffsptk 2.5.0 has been released! This update introduces differentiable WORLD (D4C and CheapTrick) modules and removes librosa dependency. github.com/sp-nitech/diff…

We would like to announce the release of SPTK 4.3. This update includes some useful new commands. We hope you find them helpful! github.com/sp-nitech/SPTK

diffsptk 2.3 has been released. This update introduces differentiable gammatone filter bank analysis and synthesis modules. github.com/sp-nitech/diff…

NEWS: diffsptk 2.0 has been released. The functional API and CQT/ICQT modules based on Librosa have been added. github.com/sp-nitech/diff…

It's that time of year again!🎄 We would like to announce the release of SPTK 4.2. A wrapped WORLD vocoder is included in this release. Please feel free to use it. github.com/sp-nitech/SPTK

We had a poster presentation on the recent progress of SPTK at SSW2023 in Grenoble. The paper is publicly available in the ISCA Archive. isca-speech.org/archive/ssw_20…

NEWS: diffsptk has been updated to 0.6.0. An inverse STFT module and vector quantization wrappers are added. github.com/sp-nitech/diff…

Subband analysis and synthesis: % x2x +sd data.short | pqmf -k 2 -m 20 | decimate -l 2 -p 2 | interpolate -l 2 -p 2 | sopr -m 2 | ipqmf -k 2 -m 20 | gwave output.png

Decomposing a waveform: % x2x +sd data.short | pqmf -k 2 -m 20 > data.sub % bcp -l 2 -s 0 -e 0 data.sub > data.sub1 % bcp -l 2 -s 1 -e 1 data.sub > data.sub2 % x2x +sd data.short | cat - data.sub1 data.sub2 | gwave -i 3 -r wave.png (top is original waveform)

Computing minimum and maximum: % nrand -l 1024 | minmax | x2x +da -2.55409 3.68321

Computing mean and variance: % nrand -l 1024 | vstat | x2x +da 0.0271787 1.02628

Drawing running spectra: % x2x +sd data.short | frame -p 80 -l 400 | window -l 400 -L 512 | mgcep -l 512 -m 24 -a 0.42 | mgc2sp -l 512 -m 24 -a 0.42 | grlogsp -g -l 512 -s 50 -e 100 -x 16 -W 500 -H 600 spec.png

Happy Christmas!🕯️ Let us announce the release of SPTK 4.1. Some signal processing commands have been added in the release. Please feel free to download it. github.com/sp-nitech/SPTK

Interpolation of F0 contour: % x2x +sd data.short | pitch -s 16 -p 80 -L 80 -H 200 -o 1 -a 1 > data.f0 % magic_intpl -magic 0 data.f0 > data.ip.f0 % cat data.f0 data.ip.f0 | fdrw -n 240 -g -y -10 200 -lw 2 -xname "Time [frame]" -yname "Fundamental frequency [Hz]" pitch.png

Converting F0 to log F0 and double its value: % x2x +sd data.short | pitch -s 16 -p 80 -L 80 -H 200 -o 2 -a 1 | sopr -magic -1e+10 -EXP -m 2 -MAGIC 0 | fdrw -g -y -10 400 -lw 2 -xname "Time [frame]" -yname "Fundamental frequency [Hz]" pitch.png

Showing 0th mel-cepstral coefficients: % x2x +sd data.short | frame -p 80 -l 400 | window -l 400 -L 512 | mgcep -l 512 -m 24 -a 0.42 | bcp -s 0 -e 0 -m 24 | x2x +da | less

Showing contents of a binary file: % x2x +sa data.short | less

Drawing a histogram of F0: % x2x +sd data.short | pitch -p 80 -s 16 -o 1 | sopr -magic 0 | histogram -l 80 -u 160 -b 20 | fdrw -b -g -xbias 80 -xscale 80 -xname "Fundamental frequency [Hz]" -yname "Count" histogram.png