A naukri.com initiative
Analyticsindiamag
1w
309
Image Credit: Analyticsindiamag
This Surat-Based Startup Builds Real-Time Speech AI Model That Cuts Costs by 20x
- Launched in 2021, Surat-based startup VideoSDK is changing how developers use real-time AI for various applications, introducing a new small language model (SLM) for businesses.
- VideoSDK aims to automate communication-intensive tasks, offering tools for embedding real-time voice and video features in applications across platforms like Android, iOS, and web.
- The startup secured a $1.2 million investment from GVFL in its last funding round, focusing on product development and go-to-market strategy, with successful implementations like Video KYC for Groww.
- VideoSDK introduces NAMO-SSLM, a real-time speech AI model combining on-device computing with cloud capabilities to reduce costs by nearly 20 times compared to other models, made language-agnostic and cost-effective.
- The model's design allows it to operate on devices like iPhones and Android phones in real time, offering privacy benefits and cost savings for applications.
- VideoSDK plans to release the model weights as an open-source initiative, aiming for community usage and scalability through a cloud offering for developers.
- Challenges faced by VideoSDK include deploying the SLM across diverse devices, especially low-end models, and improving training cycles for AI models to enhance performance.
- In the future, VideoSDK aims to expand SLM deployment across all devices, establish its presence in real-time AI communication, and achieve significant utilisation growth.
- The startup intends to compete globally with major players in the AI communication space like Agora and Twilio Video, positioning itself as a leader.
Read Full Article
18 Likes
For uninterrupted reading, download the app