Scaled Cognition Targets Enterprise CX with Ultra-Low Latency Voice Model
The landscape of enterprise customer experience (CX) is undergoing a significant shift as AI model lab Scaled Cognition announces the launch of its new text-to-speech engine, Voice. Designed specifically for high-stakes customer interactions, the model aims to eliminate the traditional trade-off between premium audio quality and operational affordability.
Bridging the Latency and Cost Gap
In the competitive arena of automated customer service, latency remains a primary obstacle to natural interaction. Scaled Cognition’s Voice model achieves a 70-millisecond speech-generation latency, ensuring that AI-driven responses occur at human-level speeds. By removing the lag that often results in awkward silences during phone or digital interactions, the company hopes to provide a more seamless experience for callers.
Beyond performance, the company is addressing the financial barriers that often prevent enterprises from deploying high-fidelity voice across all touchpoints. With a self-hosted pricing model set at $1 per million characters, Scaled Cognition is positioning itself as a cost-effective alternative to established market leaders. The API-based service is also available at $9 per million characters, aiming to democratize access to high-quality synthetic speech for large-scale operations.
Precision and Integration
A persistent challenge for text-to-speech technology is the accurate interpretation of context-heavy data. The new model features built-in text normalization, which automatically handles the complexities of dates, addresses, account numbers, and currency. This functionality allows the system to intelligently differentiate between common abbreviations and specific industry terminology, a critical requirement for sectors like banking, insurance, and healthcare.
The technology is designed for interoperability, functioning alongside existing legacy systems and current agent platforms. This modular approach allows businesses to adopt the voice engine without necessitating a complete overhaul of their broader infrastructure. Furthermore, when paired with the company’s flagship model, APT—which focuses on hallucination-free intelligence—the system ensures that the information verified by the AI is communicated with precise, reliable verbal delivery.
Market Positioning and Validation
As part of its go-to-market strategy, Scaled Cognition is emphasizing transparency by inviting prospective clients to participate in blind listening tests. By comparing their model directly against industry-standard providers, the firm seeks to prove that its solution can maintain superior audio quality while operating at a fraction of the cost. The model launches with support for over 25 languages and includes voice cloning capabilities, signaling a focus on global enterprise scalability.
Founded by AI experts Dan Klein and Dan Roth, Scaled Cognition continues to expand its footprint in the agentic AI sector. With $100 million in Series A funding led by Khosla Ventures, the company is doubling down on its mission to provide reliable, enterprise-grade AI tools that handle sensitive tasks ranging from financial transactions to complex service inquiries.
