Shunyalabs launches Zero STT Med, beating Whisper and AWS in medical ASR accuracy
When doctors need notes right away, they’ve always had to choose between speed and accuracy. It seems a new home-grown speech engine from India may have tipped the scales. In a recent test it beat both OpenAI’s Whisper and Amazon’s cloud-based service on a batch of clinical recordings, a result that could change how hospitals treat voice data.
The startup behind it, the same team that builds voice-AI infrastructure, says the model was trained on a carefully selected set of medical conversations. That lets it catch the jargon and abbreviations that generic systems usually miss. Their engineers point to a fast training pipeline that can be tweaked for specialty-specific vocabularies, and a deployment option that works on-premises or in the cloud - handy for navigating data-privacy rules.
They claim accuracy that tops the current market leaders, which feels timely as health systems scramble to digitise workflows without hurting compliance or quality. The company just released an official announcement, and I’m curious to see if the hype holds up in real-world use.
Shunyalabs.ai, a Bengaluru-based voice AI infrastructure company, has announced the launch of Zero STT Med, a domain-optimised automatic speech recognition (ASR) system designed specifically for medical and clinical workflows. The new system promises high accuracy, rapid training, and flexible deployment options, targeting hospitals, telemedicine platforms, and ambient scribe systems. According to the company, Zero STT Med achieves a word error rate (WER) of 11.1% and a character error rate (CER) of 5.1%, outperforming major ASR competitors such as OpenAI's Whisper, ElevenLabs Scribe, and AWS Transcribe. The system can be trained in just three days on 2×A100 GPUs, significantly reducing the data and compute requirements for healthcare speech models.
Zero STT Med is the newest speech-to-text model from Shunyalabs, a Bengaluru-based startup. They say the system hits an 11.1 % word error rate and also reports a character error rate, though they don’t spell out the exact number. According to the company, it beats Whisper and AWS on that same metric - but we haven’t seen any third-party tests to back that up.
The team touts a quick training cycle and a deployment that can be tweaked for hospitals, tele-medicine platforms and ambient scribe setups. If the figures are right, clinicians might see fewer transcription mistakes, which could lighten their workload and make records more reliable. Still, the lack of independent validation makes the performance claim feel a bit shaky.
Zero STT Med leans heavily on domain-specific data, so a tighter training set could be why the numbers look good. It’s unclear whether the model will hold up with varied accents, different specialties or noisy rooms. For now, Shunyalabs is pitching it as a niche-focused clinical AI tool, waiting for broader testing to prove its worth.
More in Business & Startups
AI shifts hiring: companies favor generalist engineers over specialists
Why are hiring managers rethinking their talent playbooks? The rise of AI tools is nudging firms...
Rivian showcases new AI chip and silicon inspection labs in office tour
Rivian is opening its doors to a behind‑the‑scenes look at the hardware that could define the next...
Anthropic's AI store profits as CEO acknowledges eternal transcendence
Anthropic’s new AI storefront is finally turning a profit, a milestone that feels almost ceremonial...
Gnani.ai launches Vachana STT model as core IndiaAI infrastructure
Gnani.ai, the Bengaluru‑based speech‑technology firm, has just put its newest speech‑to‑text engine...
PC Company Invests in AI Engines, Context‑Aware Computing, Intelligence
A PC maker long known for assembling laptops and desktops is now positioning itself as a backbone...
Common Questions Answered
What accuracy metrics does Zero STT Med achieve compared to Whisper and AWS?
Zero STT Med reports a word error rate (WER) of 11.1%, which the company says is lower than the WER achieved by OpenAI's Whisper and Amazon's AWS medical ASR on the same test set. The exact character error rate is not disclosed, and independent benchmarks have not been published to verify the claim.
How does Zero STT Med cater to medical and clinical workflows?
The system is a domain‑optimised automatic speech recognition (ASR) engine built specifically for medical transcription, tele‑medicine platforms, and ambient scribe setups. It promises rapid training cycles and flexible deployment options, allowing hospitals to integrate the model on‑premises or in the cloud as needed.
What are the deployment advantages of Zero STT Med for hospitals?
Shunyalabs highlights that Zero STT Med can be trained quickly and deployed flexibly, supporting both on‑site installations and cloud‑based environments. This adaptability helps hospitals choose the infrastructure that best meets their security, latency, and scalability requirements.
Who developed Zero STT Med and where is the company based?
Zero STT Med was developed by Shunyalabs.ai, a Bengaluru‑based voice AI infrastructure startup. The company specializes in building speech‑AI platforms and is now extending its portfolio to the medical domain with this new ASR model.