Image

Music


I currently know how to play 1 instrument, The keyboard (Digital piano) Keyboards are amazing instruments with the ability to mimic many other instruments and use unique synths. On the keyboard i can play, Deck the halls, Super Mario Bros. 1-1, The mario power star theme and the Wallace and Grommit theme, I am also very good at improvisation.

I also use tools such as Audacity to edit songs and make remixes, i recently made an edited version of Bohemian Rhapsody using waveform inversion to extract reversed reverb from the song.



I recently started doing some more keyboard music and am making some songs of my own as well as learning to play some more songs. I also learned how to keep in a temporal rythm and play melodies in E minor A minor.

Audacity

Using the Audio manipulation tool "Audacity" it is possible to apply various digital effects to any kind of audio as well as generating unique audio with its (rather limited) audio generation tools, Audacity is a great piece of software, it is very useful for a variety of purposes, it can also splice audio files together, and is extremely customizable as it can use a variety of audio plugins to allow for new effects to be used, I have recently made an edited version of the Playstation 1 Startup sound, as well as various other songs.




ARSS and spectrograms

I recently saw an interesting experiment using a piece of software calles Analysis Resynthesis Sound Spectrograph (ARSS), a tool which can convert any 16-bit PCM Wave file to a bitmap image and convert it back again, in the experiment ARSS was used to convert audio samples of speech into spectrograph images, where the neural network, HyperGAN was used generate new images, the idea of the experiment was to see if an AI could replicate human speech with completely unique audio patterns based on only 1.28 second clips, the results were a watery, metalic sounding muffled speech, however the speech was just similar enough to be believed to be from the person the original audio samples came from. seeing this made me interested in ARSS as i had previously wanted to try converting sound into images and back, so after seeing the experiment online, i downloaded the linux binary of ARSS and tried it out, after figuring out how to format the audio correctly, i managed to inset a 2-minute audio file and convert it to a bitmap image, the spectrograph images are limited to greyscale as colour would be much harder to process and edit, and is not needed to completely convey sound. after converting an audio file of Nintendos Mii channel theme to an image, i inspected it in the image editor, Kolourpaint, the image had been formatted to the specifications i had input, and had a resolution of 29452x115, time was portrayed horizontally with pitch shown vertically, the image shows what pitches are being played at what volume at what time, unfortunately, even though a spectrogram can theoretically portray any sound perfectly, ARSS cannot exactly reconstruct sound, after processing, the sound is always slightly off and a bit metally, although, if the resolution is good then the sound is almost perfect. After converting the sound to the spectrogram, i took the image and restriced it to purely monochrome, meaning that instead of greyscale (often called black and white), it was only the colour black and white, with no shades of grey in between, though i used automatic dithering to ensure that as much of the sound was kept intact while still allowing for some unique distortion, after that i converted the image back to an audio sample, the results were interesting and i will show them below.


© 2018,2019