aiOla Launches QUASAR, a Gateway for Hyper-Personalized Speech Recognition Routing

aiOla, a voice artificial intelligence company advancing speech recognition technology and voice agents, today launched QUASAR (Quality-Weighted Unsupervised ASR Assessment & Ranking), a speech intelligence gateway for more consistent and reliable automatic speech recognition (ASR) by dynamically routing each audio request to the ASR engine most likely to perform best for that specific moment.

With its tailored approach, QUASAR enables organizations to adapt transcription decisions in real time based on speaker characteristics, audio conditions and domain context. QUASAR acts as an intelligent gateway, evaluating each incoming audio request and routing it to the ASR option most likely to deliver the highest-quality transcription for that specific context. The platform supports orchestration across multiple ASR sources, including commercial cloud APIs, self-hosted engines and custom deployments, without forcing businesses to commit to a single provider long term.

"Until now, enterprises have been forced to choose one ASR engine and accept its blind spots," said Amir Haramaty, co-founder and president of aiOla, in a statement. "QUASAR changes that by treating speech recognition as a dynamic problem, where the best option can shift from one interaction to the next based on real conditions, not averages. This is a major leap for the industry and potentially a massive disruption for how ASRs are being consumed."

In internal evaluations across six diverse benchmark datasets, including clean read speech, professional talks, varied accents, institutional audio and domain-heavy financial content, QUASAR achieved 88.8 percent overall accuracy in selecting the best-performing ASR option, or an equivalent top choice when results were effectively tied. On clean speech, accuracy reached up to 97 percent, while on more challenging audio involving accents, noise and specialized vocabulary, accuracy remained in the 79 percent to 88 percent range.

"Voice is becoming one of the primary interfaces to technology, and that shift demands a fundamental rethinking of how speech recognition works," Haramaty said. "We're entering a new era in voice AI, one where ASR must function as living infrastructure and not a static model choice. QUASAR brings that vision to life by operationalizing speech recognition at scale, improving consistency across diverse populations and environments, enabling faster expansion into new regions and verticals and unlocking smarter cost-quality tradeoffs without vendor lock-in. The result is a platform that can transform the entire voice ecosystem, from individual developers building captioning tools to global contact centers processing billions of minutes of audio each year."