Built for the language of pharma
Power accurate, reliable, and secure pharmaceutical voice applications with speech-to-text optimized to transcribe drug names, pharmaceutical terminology, and medical vocabulary.
The leader in pharmaceutical speech recognition
General-purpose clinical STT models routinely mistranscribe drug names and pharmaceutical terminology, introducing errors that can affect patient care and critical pharmaceutical workflows. Nova-3 Pharma is specifically designed to accurately transcribe the terminology those workflows depend on.
#1 in pharmaceutical terminology accuracy
Accurately transcribes drug names, pharmaceutical terminology, and medical vocabulary while maintaining the same high-quality transcription performance of Nova-3. It ranks #1 in drug name recognition across both batch and streaming among the models evaluated.
Designed specifically for pharmaceutical workflows
Build pharmacy IVRs, prescription refill automation, member services and contact center voice agents, and clinical documentation on a single speech-to-text platform.
Seamless deployment
Deploy using model=nova-3-pharma with no infrastructure changes required. Nova-3 Pharma uses the same architecture and compute footprint as Nova-3 Medical, with support for streaming and batch transcription.
Enterprise-grade security
Flexible deployment options for regulated healthcare environments include hosted API and self-hosted deployment, with HIPAA-eligible infrastructure and Zero Data Retention (ZDR) available for enterprise deployments.
The pharmaceutical speech-to-text model you can trust
Nova-3 Pharma ranks #1 in drug name recognition and overall transcription accuracy across batch and streaming among the models evaluated.
#1 Drug name recognition
Batch + streaming
Achieves the highest drug name KRR among the models evaluated: 91.59% in batch and 91.28% in streaming. Keyword Recognition Rate (KRR) measures whether specific drug names spoken in the audio are correctly captured.
#1 Overall transcription accuracy
Batch + streaming
Delivers the lowest WER among the models evaluated: 10.17% in batch and 11.69% in streaming. Word Error Rate (WER) measures accuracy across the full transcript, with lower scores indicating fewer transcription errors.
Deepgram has been an invaluable partner for Insight Health. I evaluated a few other Speech-To-Text products, and Deepgram has the lowest latency, fewer errors, and well-documented APIs. We got HIPAA BAA signed quickly and had everything up and running in production within a week. They iterate quickly, releasing new models, features, and products. If you're an early-stage startup looking for reliable, efficient, and innovative NLP solutions, Deepgram is hard to miss.
Frequently Asked Questions
What our pharma and healthcare customers say on G2
Deepgram did a great job building a community around their speech-related AI models. Their support to the community is phenomenal, whether it's about implementation issues or resolving support tickets. We also appreciate that they have a special medical AI model, which has helped us tremendously in our venture. Lastly, their transcription is almost real-time, which is fantastic. We are very excited to try their streaming model as well. The documentation is very comprehensive and very easy to understand. We use it almost for everything related to Speech to Text.
We conducted extensive research and performance evaluations for our medical-domain speech-to-text application and determined that Deepgram's specialized models offer superior performance and latency. We also value the platform's additional features, such as speaker diarization and precise timestamping, which exhibit exceptional quality.
In pharma, every term matters
Test the medications your workflows depend on in the Playground, or talk to us about self-hosted, SageMaker, and enterprise deployment.