Smallest.ai, a leading AI company headquartered in San Francisco, has launched Lightning, the world’s fastest real-time text-to-speech model. Game-changing technology can synthesize up to 10 seconds of audio in a record time of merely 100 milliseconds, setting a new benchmark for the voicebot industry. The ultra-low latency of Lightning empowers developers to create the most realistic voice bots with instant responses while easing their integration.
The model is path-breaking on several parameters. Firstly, it allows the providers of voice bots of the whole world to create hyper-realistic audio both in English and Hindi for up to five times cheaper than the existing models. At just $0.02 per minute or approximately Rs 1.6 per minute. Lightning decreases the cost of building and running voicebots for a wide array of applications. This low cost opens ways for population-scale use cases where voicebots are now within an operational cost of less than 1 Rs per minute.
Historically, real-time text-to-speech models required complex streaming technology that increases server load and costs. Most of these systems will use a websocket connection which becomes hard to scale efficiently. Lightning removes these barriers. Meanwhile, with Lightning, real-time audio is provided within just 100ms using a simple REST API. This reduces computational overhead and allows scaling much faster while still keeping API costs much lower. That way, voicebot providers can easily scale while keeping their expenses in order.
Currently, the model supports multiple accents in English and Hindi. Though the ambitions go further into covering other Indian, European, and Asian languages over the next few months. Some of the early access voicebot platforms that have already used Lightning report an 8x reduction in costs per minute, with noticeable improvements in voice quality.
While intended for real-time applications in the first place, Lightning can support a wide array of applications. Such as audiobook creation, voice-overs for social media like Instagram and YouTube, and many other applications. For those who aren’t developers, Smallest.ai makes the technology available through its speech platform called Waves, which pushes the envelope even further with additional functions such as voice cloning and accent conversion currently in beta.
According to Sudarshan Kamath, the founder of Smallest.ai, “Why is it that 1 billion people are not speaking to AI voices every day? That is regardless of all these incredible advancements in Voice AI. So this is the problem we’re trying to solve.”
Smallest.ai was founded by IIT graduates Sudarshan Kamath and Akshat Mandloi. Prominent investors supporting Smallest.ai include 3one4 Capital, Better Capital, and Upsparks Capital. It is also on Discord, one of the largest speech AI communities in the world.
With Lightning, Smallest.ai is democratizing the voice AI space by making available quality, affordable voicebot technology at scale.
ANI
