Speaker 1: Music making is a joyous, emotional and intense experience for the composers and writers in the industry. The act of manifesting ideas and feelings into melodic and sometimes non-melodic riff-driven songs on our instruments that form the universal language of expression. A whole host of persons contribute to the entire process of putting out an album. To write music, an important step in the way is to write sheet music, once the song's structure and melody has been composed and tracked. The process of writing a tablature is very elongated, cyclic and a meticulous one. It requires very undeterred attention and a steady hand to encapsulate all the nuances of the playing within the documentation. The task itself, while of paramount significance, is mostly outsourced to expensive medias and involves persons sitting for prolonged amounts of time to complete it, sharing its nature with data entry jobs only more nuanced. Now that is where strings come in. The major problem within the process of writing or documenting music that strings seek to eliminate is the redundant amount of labour, time and money spent in the task. Music transcription of some audio signal essentially refers to the process of extracting digital data of the corresponding audio waveform and using it to obtain the symbolic information associated with the more higher level music notation structures and then representing it as it would be on a music score sheet. Information regarding an audio signal such as the BPM, Timbre, Tone, Frequency and Amplitude are absolutely essential for the purpose of achieving a relatively accurate deduction of tablature for the audio slash musical piece. Having barely any idea of what we were getting ourselves into with the project, our humble beginnings started from googling as much as we could about sound and audio analysis. Our research took us to several avenues within the field which looked promising to say the least. Nevertheless, the field of study we undertook was still very nascent. Our research did however help us to understand the underlying concepts of music theory and led us to the primary algorithm of use, Fourier Transform, the serviceable idea that would help us isolate the various natural frequencies or notes in the universal chromatic system of music from the audio sample of our choosing. Our methodology and workflow behind the project were an oscillation between coding an approach we read about and scratching it to spend more time researching a better method for almost the entirety of this project. The coding was slow, but the learning experience involved in it was a treat to aspiring CSE undergrads like us. The research papers on what we were trying to accomplish were as new as days behind with PG students from Ivy institutions publishing their approaches for the same. To work on technology of such great significance and having active community background was truly humbling and exciting. But it also meant that we'd have to keep changing our approach every now and then to be able to accomplish our goals and procure deliverables. After sustained work and research on the project for the first few months, we were able to boil down the data flow of the system to an initial supposed structure which has been subject to minimal changes since, owing to the nature of our workflow and the topic of the project. Instrumentationalists will record composed pieces of their instruments into their computers during a process called as tracking. Tracking may be done through various processes such as using a virtual DI or digital input in which the instrument is directly plugged in through an interface and takes MIDI input or they may mic up the external amplifier for capturing a more raw tone or sound. In addition of being able to use real-time analysis, we may also browse and locate a pre-recorded audio file for the same purpose. The file thereafter will go for layer analysis and hidden layer detection which would happen frame by frame and would be useful for the next procedure of layer separation. All of this will precede the process of Fourier transform. Hereafter, Debian models may be used for classification and attaining a segmented audio sample. This segmented audio sample can be driven through three processes for three different results. Firstly, we can either drive it through the discrete Fourier transform which will give us a Mel-Felder band and then after a few procedures, we will be able to obtain a Mel spectrogram that gives us some relevant information regarding the audio sample. The second method that we may follow would be a fast Fourier transform where after through processes such as pitch class profiling and smoothening of convolutions, we will have achieved a smoothened PCB which would help us obtain the note detection process wherein we will be able to detect whatever note is being played in the audio sample. The third and the final process that we may perform with the segmented audio sample that we have previously obtained would be key detection and enhanced beat detection which goes through enhanced correlation and pitch class profiling. For key detection, we may use some reference pitch class profiles and then use them for comparison. Upon comparison, we can obtain a assumed chord sequence based on probability which thereafter will be optimized to give us the final chords that are being used within the audio sample that has been provided. So, all in all, that has been our concept for Strings, the automatic music transcription software that our team of dedicated and enthusiastic undergrads have meticulously persevered to perfect. We hope to deliver on everything that Strings hopes to achieve through further continued research. Our team looks forward to putting Strings out as a free-of-cost software to contribute to the music industry in a positive way. And we'd like to thank our college Bennett University's computer science department and our teachers for giving us the opportunity and resources to work on such an amazing project. Thank you.