Bengaluru-based AI company Sarvam AI has introduced Bulbul V4 Flash, a text-to-speech model designed to generate speech in 11 languages with a time-to-first-audio latency of less than 250 milliseconds. The launch aims to make real-time voice applications faster and more affordable.
The model is priced at ₹20 per 10,000 characters, down from ₹30 for Bulbul V3, representing a reduction of approximately one-third.
Sarvam AI said the new model improves pronunciation, pacing and contextual delivery while reducing the time and cost involved in speech generation. It is designed for real-time applications such as customer service and AI-powered voice agents.
“With V4 Flash, our research focused on capturing the richness and nuances of Indian languages, while making high-quality speech generation faster and more affordable,” the company said in its announcement.
Bulbul V4 Flash is available through the Sarvam API, Playground and Content Studio. It offers 83 voices and 224 voice-and-style combinations for use across banking, e-commerce, education, sales, customer support and narration. According to the company, the model is up to 40% faster than Bulbul V3 under comparable workloads. Its infrastructure has also been optimised to support multiple conversations simultaneously. Existing users can migrate through the current Bulbul API.
To improve speech quality, Sarvam AI expanded its proprietary speech datasets to nearly 7 times the size used for V3. The expanded datasets cover more languages, regional accents, speakers, emotional expressions and conversational styles.
The company pretrained V4 from scratch, focusing on better representation of Indian languages through a language-balanced training process. This approach aims to improve pronunciation while preserving the distinctive characteristics of each language.
“Regional voices sound native, Indian English sounds Indian, and names, numbers and everyday expressions come out the way people say them,” the company said.
The model also features a contextual text normalisation layer that determines how names, dates, amounts and abbreviations should be spoken. It can distinguish between abbreviations pronounced as words and those spoken letter by letter. This is particularly useful for voice agents handling customer details, where errors in names, payment amounts or booking references can cause confusion.
Sarvam said its research explored larger language backbones, different audio representations and generative approaches, with training data scaled to more than 400 billion audio tokens. These experiments helped the company develop a smaller model that retains key speech-quality improvements while meeting production requirements for speed and cost.
Its streaming-first inference system prioritises the next audio segment instead of generating speech far ahead of a conversation. The company has also reduced unnecessary synchronisation and data copying to improve performance during concurrent requests.
Separately, Sarvam previewed Bulbul V4 Pro, a model under development for more expressive speech. The preview demonstrated emotions such as happiness, sadness, fear and anger, along with actions including laughing, sighing, whispering and emphasising specific words. Sarvam has not announced its release date or pricing.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.