Expertise has at all times been an ideal leveller. From the economic age to the age of the internet, it has improved the standard of life for the plenty and made issues beforehand unimaginable extra accessible. One solely wants to try their smartphone to know how speaking with somebody sitting hundreds of kilometres away has change into so frequent that the majority don’t even give it some thought a lot. Earlier than Graham Bell, such long-distance communication was solely obtainable to the wealthy and influential as a result of excessive prices related to it.
Such examples are numerous. From social media offering true connectivity internationally, smartphone apps digitising duties that required bodily presence and took away hours from a day, and distant work that empowers individuals residing distant from company hubs with higher incomes alternatives, expertise has democratised accessibility itself. In some ways, generative artificial intelligence (AI) has change into the following torch-bearer to increase accessibility to new frontiers.
One such space the place accessibility could make a huge impact is the music business. Regardless of the arrival of impartial streaming platforms equivalent to Spotify, SoundCloud, Apple Music, and extra, making music distribution cheaper, the issue assertion that is still is music creation. At present, authentic background music is a much-needed commodity. From skilled artists to social media creators and podcasters, everybody requires music tracks for his or her content material, ideally authentic, to keep away from any copyright strikes by platforms (YouTube content material creators are effectively conscious of its impact) or a lawsuit.
However creating music is just not everybody’s cup of tea. Probably, when you’ve got not skilled for years to grasp one or a number of musical devices, but you need authentic and distinctive music to your skilled wants, you end up caught with solely two costly options — rent a music producer or a session musician, or pay on-line to purchase inventory music. However not anymore, as a result of that is the place AI has stepped in.
Take the instance of Beatoven.ai, an Indian AI-powered music technology platform that lets customers write a easy textual content immediate to generate new and distinctive background music inside ten seconds. To grasp how this expertise works, its varied implications, and the expertise of operating such an revolutionary startup, we at Devices 360 spoke with Mansoor Rahimat Khan, the co-founder and CEO of Beatoven.ai.
The inception and journey of Beatoven.ai
Mansoor Rahimat Khan, CEO and cofounder of Beatoven.ai on the Devices 360 Awards
Mansoor Rahimat Khan comes from the Gwalior-Indore-Dharwad Gharana of Sitar, a well-known household of musicians which have performed and formed modern-day Sitar music for seven generations. Khan was no completely different, however he selected a special path owing to a different of his passions — expertise. “I accomplished my commencement from the Nationwide Institute of Expertise (NIT), Goa, in electronics and communication engineering. This was additionally after I began delving into the area that lies on the intersection of music and expertise,” Khan instructed us.
After working for just a few years, Khan met Siddharth Bhardwaj, an alumnus of the Indian Institute of Expertise (IIT), Allahabad (now referred to as Prayagraj), and a music fanatic. The duo, sharing comparable pursuits, recognized the issue of music licensing in content material and needed to construct one thing that would make music extra accessible to thousands and thousands of creators — whether or not on social media or professionally pursuing a profession. That was the genesis of Beatoven.ai.
However there was one downside. Even because the duo started engaged on the product and the startup in 2021, their answer to the issue required generative AI, which was nonetheless a yr away from reaching the mainstream (in November 2022, ChatGPT arguably began the gen AI race).
“Initially, the prototype we in-built 2021 was a really bare-bone platform. Customers might choose a style and a tempo and specify a period, and we might generate an authentic piece of music. Again then, no massive language fashions (LLM) existed, so we needed to construct our whole tech stack from scratch. At present, we now have our personal proprietary tech that we began constructing again then,” Khan stated.
Issues grew to become simpler as soon as the AI wave took place, and Beatoven.ai benefitted from the supply of LLMs available in the market, utilizing which they might higher equip their platform to cater to its present person base of 1 million.
The Beatoven.ai platform
The net-only platform is a generative AI-powered music technology device for content material creators. Customers, as soon as they’ve signed up, can write a textual content immediate to generate authentic background music. Alternatively, the platform additionally permits customers to select a tempo, period, style, and temper to create music.
As soon as the person has added the enter, the AI takes over and generates 4 separate tracks. The platform additionally provides post-generation modifying options the place customers can change an instrument, scale back or improve quantity in particular components, or recompose a complete part of the monitor. Khan stated a single monitor could be as much as quarter-hour lengthy, though there is no such thing as a higher restrict, and the urged worth exists to maintain rendering time quick. A monitor of a mean size of 1-2 minutes will take about 10 seconds to generate. Primarily based on information shared by the corporate, since inception, Beatoven has generated 15 lakh soundtracks and boasts 3 lakh downloads.
The platform at present doesn’t permit customers to make fusion tracks the place two or extra genres are blended, however Khan instructed Devices 360 completely that the corporate will quickly launch a brand new replace that can add this function.
We additionally examined out the platform and located the music to be fairly sensible. The next track was created utilizing the immediate “Create a high-energy EDM anthem with a beat drop that’s excellent for a dance social gathering”.
The Beatoven.ai tech-stack
There are two parts to the Beatoven platform. The primary is the LLM, which permits customers to kind prompts in pure language after which course of that data in a format the AI can perceive to transform it into music. The startup makes use of GPT fashions for this half.
The second element understands the person intent and generates a monitor that fulfils the parameter. This structure was created by the corporate natively. The AI mannequin makes use of contrastive studying structure to make it occur. Khan highlights that the inspiration for this method got here from OpenAI’s CLIP mannequin, however shortly factors out that the OpenAI mannequin was constructed for textual content and pictures, and Beatoven was the primary to make use of it for sound and music. As a consequence of it being a proprietary work, the corporate was additionally capable of optimise the method. For example, Khan instructed us that the platform makes use of CPU inference as an alternative of GPU inference. That is notable given even small LLMs require GPU inference to run.
The startup has sourced nearly 1,00,000 information samples from impartial artists to coach the AI mannequin. The corporate collaborated with practically 250 artists globally and paid them for unique tracks. Khan claimed that the corporate had ethically sourced all of its coaching information and didn’t scour the web for it. Apparently, Adobe is reportedly doing the identical at current to construct an AI video technology mannequin.
Nevertheless, information, at the moment, has change into an extremely pricey useful resource that’s required regularly to improve AI fashions and enhance them. Whereas Beatoven continues its follow of collaborating with artists to acquire information even at the moment, sooner or later, it plans to chop prices by introducing a revenue-sharing mannequin, the place artists can be paid based mostly on the variety of tracks generated the place the AI used the track pattern or the information.
How Beatoven.ai plans to take care of the competitors
AI-based music technology is just not totally a singular proposition at the moment. Many gamers have entered the section, recognising the potential. Some embody Google with its MusicLM, OpenAI with its Jukebox, and Adobe with its Challenge Music GenAI Management. Nevertheless, none of those fashions is obtainable to the general public at the moment, and so they stay underneath growth. However competitors for Beatoven nonetheless exists. A giant rival for them can be Suno AI, which not solely creates music but in addition provides AI-generated voices to the music to supply a full-fledged track.
In reply to the priority, Khan highlighted that the corporate provides limitless music technology with out including a price restrict. Additional, he highlights that the corporate is constructing a complete ecosystem. Whereas on one facet, it’s catering to customers by producing music, alternatively, it additionally provides a spot for artists to promote their authentic music. The whole suite of choices, together with the promise of “ethically sourced and copyright-free distinctive music”, is what Khan believes offers Beatoven the sting available in the market.
A glance in direction of the longer term
Beatoven is now taking a look at enlargement of its platform to cater to a world person base. The startup has already begun onboarding artists from completely different components of the world as 70 % of its person base resides exterior the nation. Khan believes this world outlook, together with specializing in bettering the AI mannequin, would be the key to hitting its goal of 5 million customers within the subsequent two years.
Expertise can usually be a two-edged sword. Whereas the advantages of AI-generated music can’t be understated, the query that arises is whether or not such simple and reasonably priced music creation can have an hostile influence on aspiring musicians. Is the commodification of music actually the correct strategy to go?
Khan believes whereas music creation goes to change into the following large disruption within the business, it’s unlikely to remove the goals and livelihood of musicians and singers. “I consider artists are nonetheless going to be on the centre of this disruption as a result of AI can’t compete with human creativity,” he stated.
About The Author
