When Siddharth Bhardwaj and Mansoor Rahimat Khan started working on creating an AI-powered music creation tool, they always thought of helping creators create original background music without having to deal with licensing. “We knew about the problem and wanted to solve it using technology,” says Bhardwaj, who started Bangalore-based Beatoven.ai together with Khan in 2021.
Bhardwaj, 31, always wanted to merge music with technology, so after graduating from IIIT-Allahabad, he got a Master’s degree from Music Technology Group, UPF, Barcelona, Spain. Having previously worked with startups specializing in signal processing, deep learning, and music technology, Bhardwaj understood artificial intelligence and its applicability to music generation. Starting an AI-focused company of his own was a natural progression after Bhardwaj met Khan, a seventh-generation sitar player from Dharwad gharana.
Music licenses and the copyright questions that come with them can make any creator nervous, especially someone just entering the field. Do you have the right to use this background music in your podcast? How long does copyright last? What does royalty free mean? That’s where this AI-powered music creation tool starts to make sense: it simplifies the way you create original, royalty-free background music in videos and podcasts and helps creators.
Bhardwaj says the music generation AI tool works like any music director would. Essentially, you set the length of the piece of music you want, then choose things like tempo and genre, and leave it to AI, which parses the database and gives you five pieces of music that are in sync with the theme of the video. . or podcast. You can then publish and launch the video or podcast wherever you want. If necessary, he can customize the sound and modify the instrument layers.
The system works using deep learning networks, a type of AI that analyzes large amounts of data. “The AI organizes all those loops and samples that we’ve gotten from the artists… We train our AI-based models to generate a structure that combines all these layers vertically and horizontally, to create a cohesive track,” explains Bhardwaj. . “For a one-minute clue, it will take the AI 10 seconds to generate five options for you.”
AI Music will explode, as OpenAI’s ChatGPT is taking over the world.
Bhardwaj says his company works with hundreds of musicians creating a 10-second guitar riff or 15-second piano melody for the platform. The Beatoven.ai database now has 50,000 such samples, coming from 200 artists.
“It’s not a tool for musicians,” Bhardwaj clarifies. “You can’t load any kind of sound into it…it’s more of an instrumental background music tool for videos or podcasts.”
“It’s not like you could train the AI and it would create a good piece of music,” he says. “The biggest problem with music is that copyright is strictly enforced, unlike text or images,” explains why it’s nearly impossible to take historical music data and train an AI model. “Nobody can do that. That’s why we even chose our own data set.”
AI can teach music composition, help freelancers mix and master, and even help artists with idea generation, but it cannot replace musicians. Many recording artists have experimented with AI for lyric generation, the most famous being David Bowie with the song Hallo Spaceboy. In 2018, Francois Pachet, a musician and tech researcher, released the first AI-composed pop album, Hello, World.
“For a creative endeavor like music, it will be a collaboration between a human being and AI,” says Bhardwaj. “AI is already creating something for you, but it’s your job to curate and customize it to meet your needs. It cannot be a generic tool for everyone, but it must be very personalized”, he adds.
Bhardwaj believes that AI is not only profitable, but has the potential to increase engagement and enhance the culture of storytelling with compelling background music. “Everything you hear on our platform is not synthetic or AI-generated, but it is always human-composed music,” he said.
Beatoven.ai is a Bangalore-based startup, co-founded by Siddharth Bhardwaj, who is also the CTO of the company.
Bhardwaj’s company makes money through a subscription-based model, with the basic tier costing $20 a month and the higher tier costing $100 a month. There are more than 28,000 users and YouTubers make up the majority. But Bhardwaj says that he has seen people using the tool to create background scores that will also be used in video games.
Demand for royalty-free music in India is huge, says Bhardwaj, and searches for such AI tools are estimated to be between 5-6 lahks. To convert more users, Bhardwaj’s company is accelerating its efforts to focus on regional Indian classical and royalty-free music for which he is collecting data and working on building AI models. In the coming months, Bhardwaj plans to release another AI-based tool that analyzes the video and reads the scene changes and mood of the scene and tries to deliver a perfect piece of music for it.