Why We Need to Talk About Innovation in India – Reflections from IEEE ICASSP 2025

Reflections on ICASSP 2025 in Hyderabad, highlighting India’s innovation potential and CAMB.AI’s unveiling of its groundbreaking multilingual voice AI, MARS6.
April 22, 2025
3 Minuten

Earlier this month, our CTO, Akshat Prakash,  attended the 50th International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2025) in Hyderabad, India. This milestone event marked the first time ICASSP was held in India, bringing together a global community of researchers, industry leaders, and innovators to celebrate advancements in signal processing.

ICASSP 2025 was more than just a conference; it was a celebration of five decades of progress in signal processing. Hosted at the Hyderabad International Convention Centre from April 6 to 11, the event featured a rich tapestry of technical sessions, keynote speeches, workshops, and cultural showcases, all under the theme "Celebrating Signal Processing"

A key point Akshat addressed was the perception that Indian startups often prioritise repackaging existing technologies over original innovation. He emphasised the importance of grounding products in robust research and development, stating:

"There's a lot of chatter about Indian startups not doing enough innovation and just repackaging what they see in other countries and trying to sell it... more selling, not research-based."

Being one of the very few (if not the only) companies present at ICAASP, with a business dedicated to the sports industry speaks volumes about our commitment to research and innovation. Our advanced multilingual voice AI solutions are scientifically designed to overcome localisation challenges, a capability we proudly showcased at ICAASP.

At the event, our research team, led by CTO Akshat Prakash, unveiled MARS6: our proprietary voice model. This groundbreaking 70 million-parameter hierarchical-codec text-to-speech (TTS) system is built for voice authenticity and boasts state-of-the-art voice-cloning capabilities. Supporting over 150 languages and dialects, MARS6 was the centerpiece of our research paper, “MARS6: A Small and Robust Hierarchical-Codec Text-to-Speech Model.”

Its compact size and impressive performance proves that innovation, not scale, is the key to progress. This breakthrough technology is making AI Voice Translation globally accessible, making languages an important part of infrastructure and growth.

Looking Ahead

IEEE ICASSP 2025 Hyderabad, India was a testament to India's growing influence in the global tech landscape. The discussions and collaborations that took place have inspired us to continue pushing the boundaries of what's possible in AI and signal processing.

We're excited to integrate the insights gained from ICASSP 2025 into our ongoing projects and look forward to contributing to the vibrant ecosystem of innovation that the conference so vividly showcased.

Verwandte Artikel

Voice Cloning Use Cases 2025 | AI Voice Replication Applications
January 28, 2026
3 Minuten
5 Real-World Use Cases for Voice Cloning
5 real-world voice cloning applications from film dubbing to accessibility. Learn how AI voice replication solves production challenges.
Artikel lesen →
10 TTS Use Cases 2026 | Text to Speech for Media & Apps
January 28, 2026
3 Minuten
10 Practical Use Cases for Text to Speech in Media and Voice-Powered Apps
10 practical text-to-speech applications from accessibility tools to GPS navigation. Learn which MARS8 model fits each production use case.
Artikel lesen →
January 26, 2026
3 Minuten
How to Automate Multilingual Customer Support with AI Voices
Complete guide to automating multilingual customer support with AI voices. Learn the workflow from speech capture to voice response across languages.
Artikel lesen →