Bengaluru, May 7, 2025 — Sarvam AI, a Bengaluru-based generative AI startup, has officially launched Bulbul V2, the latest version of its cutting-edge text-to-speech (TTS) model. Designed to deliver natural, human-like speech in 11 Indian languages, Bulbul V2 is positioned as a strong competitor to global TTS leader ElevenLabs, offering superior performance, lower cost, and regionally authentic voice output.
With this release, Sarvam AI aims to make voice technology more accessible, customisable, and culturally relevant for the diverse Indian market. The startup asserts that Bulbul V2 not only speaks more Indian languages than ElevenLabs, but also offers speech in authentic regional accents that sound natural, not robotic or rehearsed.
What’s New in Bulbul V2?
The second generation of the Bulbul model builds significantly on the original version launched in August 2023. Key upgrades include:
- Support for 11 Indian languages: Hindi, Tamil, Telugu, Marathi, Bengali, Punjabi, Odia, Kannada, Malayalam, Gujarati, and Oriya.
- Authentic regional accents: Voices reflect the unique phonetic and tonal characteristics of various Indian regions, enhancing relatability and realism.
- Low latency: Bulbul V2 boasts P90 latency speeds of just 0.398 seconds, a substantial improvement over ElevenLabs’ 0.945 seconds, ensuring faster response times.
- Affordable pricing: Sarvam offers its service at Rs 15 per 10,000 characters, making it up to five times cheaperthan ElevenLabs.
- Customisable voice options: Users can choose from six unique voice personalitiestailored for different industries and communication styles.
“We speak more Indian languages than ElevenLabs,” Sarvam said in a recent post on X (formerly Twitter), highlighting its commitment to linguistic inclusivity.
Six Distinct Voice Personalities for Versatile Applications
To cater to diverse industry needs, Sarvam AI offers six distinct AI voice personas:
- Amartya– Expressive and distinct, ideal for storytelling.
- Pavitra– Dramatic and engaging, great for advertisements and theatre.
- Amol– Narrational and mature, suited for documentaries.
- Maitreyee– Informative and engaging, useful for education.
- Arvind– Conversational and articulate, perfect for customer service.
- Meera– Professional and articulate, designed for corporate use.
These voice options help businesses build unique auditory branding and maintain consistent voice identity across customer touchpoints.
Behind the Scenes: Data-Driven Development
According to Sarvam’s official blog, Bulbul V2 was trained using high-quality, diverse audio datasets featuring multiple speakers and languages. The training data included code-mixed inputs, proper nouns, abbreviations, and a mix of conversational and professional tones to ensure versatility across use cases—from e-learning to enterprise voice bots.
Sarvam AI: Building India’s First Homegrown AI Foundation Model
The launch of Bulbul V2 comes shortly after Sarvam AI’s major milestone: being selected by the Government of India as the first startup to build the country’s indigenous AI foundational model. Out of 67 applicants, Sarvam stood out, earning governmental support in terms of compute resources to develop a homegrown large language model (LLM) from the ground up.
Who Is Behind Sarvam AI?
Founded in July 2023 by Vivek Raghavan and Pratyush Kumar, both alumni of AI4Bharat—an AI initiative backed by Infosys co-founder Nandan Nilekani—Sarvam AI is on a mission to democratize access to generative AI tools for Indian enterprises and the public sector.
Raghavan, a seasoned technologist who played a key role in building Digital Public Goods (DPGs) like Aadhaar, believes in creating population-scale impact by integrating generative AI into India’s digital infrastructure.
“Our goal is to co-develop domain-specific AI models in partnership with Indian enterprises, leveraging their data to solve real-world problems,” Raghavan shared.
Sarvam’s full-stack AI platform spans everything from research-driven innovations in AI training to an enterprise-grade deployment suite, making it a comprehensive solution for Indian businesses venturing into generative AI.
Strong Backing and Future Vision
In December 2023, Sarvam AI raised $41 million in a Series A funding round led by Lightspeed Venture Partners, with participation from Peak XV Partners and Khosla Ventures. This significant investment underscores investor confidence in Sarvam’s vision to create AI solutions that resonate with the Indian context and beyond.
Outlook
With Bulbul V2, Sarvam AI is not just offering another text-to-speech tool—it’s redefining the standard for voice AI in India. By combining linguistic diversity, regional authenticity, and technological excellence, Sarvam AI is taking bold steps to challenge global players and lead the voice tech revolution from India.
As the company gears up to build India’s first indigenous large language model, all eyes are now on how Sarvam will continue to shape the country’s AI-first future.
Read more: Khalifa A. AlJaziri: The Architect Behind UAE’s Smart Home Revolution