Digital health platform Eka Care has unveiled Parrotlet v2, an India focused clinical grade automatic speech recognition model designed specifically for doctors who need fast, accurate and structured medical documentation in busy OPD and hospital settings. The new real-time engine is built for Hindi and Indian English clinical speech and is positioned as a practical tool for everyday consultations rather than studio quiet recordings.
The company says Parrotlet, a v2, sets a new benchmark for AI-driven medical transcription in Indian healthcare by aligning tightly with the linguistic, acoustic and workflow realities of clinics and hospitals. As electronic medical records continue to expand across the country, clinicians are spending more time entering structured data, often cutting into the limited minutes they have with each patient. Eka Care is pitching this model as a way to convert live doctor-patient conversations into structured clinical notes within seconds, so that documentation becomes an integrated part of the consultation instead of a separate administrative load.
In high volume OPDs, where queues are long and consultations are short, manual note-taking and post-visit typing can lead to bottlenecks, delays and professional fatigue. Parrotlet a v2 is designed to sit inside that environment, listening in real time and generating usable records on the fly. The model supports clinical workflows that demand a five to ten second documentation window, allowing doctors to keep eye contact and conversation going while the system captures key medical details in the background.
“India’s clinical reality is multilingual, acoustically noisy, and filled with hyperlocal medical terminology that global AI systems are not trained to handle,” said Vikalp Sahni, Founder and CEO of Eka Care. “Parrotlet a v2 is tuned for Indian healthcare. In benchmark evaluations against leading global and India focused models, including Gemini 3 Pro, Gemini 3 Flash and Saaras V3, it demonstrated strong performance on semantic word error rate while still delivering sub-second response times during consultations.”
The model sits at the core of EkaScribe, the company’s AI driven digital medical scribe platform, which is already in use with more than three thousand doctors across specialties. With a compact architecture of five billion parameters, Parrotlet a v2 aims to balance speed, cost and capability, matching or surpassing larger models on Indian clinical speech without the same infrastructure demands. Eka Care highlights that the system can interpret code mixed Hindi and Indian English, overlapping speech, regional accents and India specific medicine names while keeping hallucinations low, so it avoids inserting details that were never spoken by the doctor or patient.
On internal benchmarking using proprietary real world doctor patient audio, Parrotlet a v2 delivered close to ninety three percent accuracy on medical keywords in Indian English and around eighty five percent in Hindi, indicating reliable capture of critical items such as drugs, doses, symptoms and diagnostic terms. The team also tested performance on noise only audio to measure phantom speech, where a model mistakes background sounds for spoken words. Parrotlet a v2 recorded a phantom speech rate of three percent, a key safety factor inside crowded hospitals where alarms, conversations and corridor sounds are constant.
Eka Care notes that while some large general purpose AI models can take up to thirty seconds to process a thirty second clinical recording, Parrotlet a v2 reaches similar accuracy with sub second inference. This translates into as much as thirty times faster performance in live workflows, a difference that can decide whether speech based documentation is viable in a packed OPD or simply too slow to be useful. The company presents this speed advantage as central to adoption, alongside accuracy and reliability.
“Healthcare digitisation at scale requires AI that is not only accurate, but fast, reliable and economically sustainable,” said Deepak Tuli, Co-founder and COO of Eka Care. “Our focus was to build a specialised AI engine that truly understands Indian clinical practice so providers can deploy it confidently across networks without compromising workflow speed, affordability or patient safety.”
With Parrotlet v2, Eka Care wants to support nationwide health system modernisation by cutting administrative load and making structured data capture at the point of care a routine part of practice. The company plans to continue its India-centric health AI research, extend language and dialect coverage and deepen links with hospital information systems, insurers and public health platforms. The stated goal is a more connected and data-driven ecosystem where clinical information flows securely and usefully across care settings.
Parrotlet a v2 is now available through Eka Care developer APIs, opening the door for integration into third-party healthcare platforms, OPD software and clinical applications that want to embed Indian clinical speech-to-text capabilities directly into their tools.
