June 10 was an unbelievable day for classical musician and software engineer Sandeep Ranade. In 2019 when he attended the worldwide developer conference in California hosted by Apple, he never thought that we would ever win in this conference. Now in 2021 his application NaadSadhana, an artificial intelligence-powered Indian classical music app has won the worldwide developers conference of 2021. When the competition kicked off his AI classical music app was featured in the best innovation category alongside league of legends: wild rift, a game by Riot Games. For Sandeep the week has been full of ups and downs with happiness and disbelief pouring in altogether.
He later got a call from Apple Bangalore who wanted to discuss further the app, the UI team helped Sandeep In correcting some technical mistakes. He even got to do a demo in front of Mike Stern who is top brass of Apple.
What is NaadSadana?
NaadSadhana was first launched in April 2018. It was an outcome of sincere help that Sandeep wanted for his friend who was learning music and needed help vocalising. Since the notes were not hitting the scale, hunt for the right music app began.
Sandeep says some apps can figure out a bad tune in any musical instrument but recognising human voice and check if it’s in tune, is not an easy task for a machine.
But he wanted to help his friend and others as well, in a month he designed a tuner tool and later started working on Indian style tuning for Hindustani and classical music in that tuner. The reception from his friends encouraged him to put this app on App Store and he named that tuner tool NaadSadhana. Ranade being a musician was always on the lookout for an app that would help him sing and maintain an ambience of concerts and practice sessions. He recalled that while he was tuningIndian classical instruments like Swaramandal which is a musical instrument with 40 strings, it was a painful task.
Everything had to be tuned to a particular Raag which requires a certain pitch (of voice or instrument), which changes from Raag to Raag. There is a total of 83 types of ragas in Indian classical music. He also tuned in Tabla and made a separate Tabla button.
When a person is singing a particular raga the notes in the melody cannot go higher than the limit of the raga. Hence it requires constant contextual analysis and artificial intelligence needs to be of top-notch for this job.
NaadSadhana also includes different harmonies and different music genres such as western, Bollywood, semiclassical and fusion. These genres require automatic chords and back tracks of piano, harmonium, violin et cetera. If a song is sung freestyle the AI won’t adhere to any raga, the singer is free to scale notes to any point. But if the music is sung in a particular raga the app has to make sure correct playback whilst analysing the context (raga) simultaneously, so that notes are correct.
App also allows recording 12 tracks simultaneously, it is called the multitrack recording feature. The app also sports a mini mixing studio for easy publishing. Once a musician starts messing with percussion instruments, multiple ragas and taals he can spend his entire life with that app and never get bored.
Sandeep used SwiftUI to rewrite the application. His experimental release of two taals for Ghazal singers; Ghazal Dadra and Ghazal Keherwa, required next-level AI. There were already 30 to 40 pieces of AI for each instrument but the demand was for a manager/conductor AI, as a concertmaster. This AI ensures that everything plays in tune even when everything is turned on.
Who is Sandeep Ranade?
Sandeep Ranade is a software engineer and a musician based in Pune. He comes from Mewati Gharana of music and is a disciple of late Pandit Jasraj, he has been trained by Shobha Abhyankar and Anjali Joglekar-Ponkshe. Mewati Gharana was founded in 19 century at Holkar court in Indore by the Khan brothers ustad Ghagge Nazir and Ustad Wahid Khan. Sandeep has been performing since the age of six and has received worldwide attention on his song ‘na Corona karo’ which was sung in Raag Basant. He has toured for music concerts extensively through North America and India and is known for the lustrous quality of his voice. He has been praised by top musicians of India and worldwide making him a world-class technocrat and a musician. The musical background and extensive training in classical music of Sandeep Ranade helped him make an app that is above the ordinary. As discussed before Indian classical music has certain notes and scales that cannot be crossed while singing a raga.
Why is NaadSadhana released only on iOS?
The app was only released on the iOS platform because of two reasons. Apple gives easy access to hardware-level complex building blocks which helps the developer to build complex structures like high-volume apps in a relatively low time. The other reason is the low latency capability of iOS and high analytical and computing speed, iOS can process 25 parallel tracks at high resolution and high speed. It also ensures that the musical instruments played feels real and not dragged. Android, on the other hand, has a higher latency of audio which is 20 to 200 milliseconds, if the instrument is delayed even by 1 millisecond the total composition of sound changes and it feels like noise. Low latency in layman’s language means minimum delay while analysing and computing a high volume of data.
What is AI in music?
The advent of artificial intelligence in music began in 1951 when British mathematician Alan Turing created computer-generated music. The machine created the melody of God save the King and Ba Ba Black sheep. Artificial intelligence is now ingrained into music-making, a common example is a software like logic which is used by musicians everywhere to create sounds of drum patterns repeatedly. Music and AI work by analysing a large number of data musicians feed the software, in terms of music right from disco, classical, dance hits, sad songs, R&B et cetera. The machine then analyses tempo, cords, notes et cetera and rewrites its melody. Machines create these outputs by either examination of data or on hard-coded rules of musical theories like Raag, taal etc.
People are making money by providing AI-powered services for creating, processing and analysing music. Some applications are
Flow machines by Sony CSI – this machine aims to generate music on its own and with human help. They aim to augment the creativity of an artist
Amper music – it helps users create music in different genres. It can be for music videos movies video games et cetera it has AI that composes, produces and perform music and video content. The web application chooses mood, style, length.
Google Magenta – the algorithms generate songs, images, drawings et cetera. Magenta was started by research engineers from the Google brain team, the software explores smart tools and interfaces for maximum benefit.
AIVA – this software enables composers to make original music and upload it further, creating new variations. The software also produces variations of existing songs.
Even after smart coded algorithms and ever-expanding machine learning the music created by these software needs a human touch. They are not award worthy music yet. But we never know what the future holds for machines and musicians. Continuous research and development in machine learning and artificial intelligence may one day eliminate the need for human musicians.