MediScript AI - AI-Powered Clinical Documentation Generator
MediScript AI is an intelligent medical transcription platform that captures patient-doctor conversations and uses generative AI to automatically create structured SOAP notes, extract
medical keywords, and suggest relevant billing codes.
Developed as an open-source blueprint under the Cloud2 Labs Innovation Hub, MediScript AI
demonstrates how Speech-to-Text models and Large Language Models (LLMs) can be
combined into a practical, ephemeral clinical documentation workflow.
It showcases a production-style architecture for capturing audio in the browser,
performing contextual speaker diarization, extracting medical entities, and generating editable, exportable clinical notes without permanently storing sensitive patient data.
What It Demonstrates
MediScript AI illustrates how to:
- Capture browser audio or upload files with built-in duration limits for clinical processing
- Convert speech to text with precise timestamps using OpenAI Whisper
- Perform contextual speaker diarization (Doctor vs. Patient) using LLM reasoning
- Generate structured SOAP notes (Chief Complaint, Symptoms, Assessment,
Recommendation)
- Extract and dynamically color-code medical keywords (Symptoms, Medications,
Diagnoses)
- Suggest automated Medical Billing codes (CPT and ICD-10) based on clinical context
- Provide an interactive "human-in-the-loop" UI for doctors to edit AI-generated notes
- Ensure compliance with an ephemeral, zero-storage backend architecture
Designed for healthcare providers, telemedicine platforms, clinical documentation
assistants, and health-tech innovation groups, MediScript AI serves as a reference
implementation for AI-driven clinical scribing.
Key Capabilities
Automated SOAP Notes
Translates raw, conversational transcripts into cleanly formatted Chief Complaint, Symptoms (as bulleted lists), Assessment, and Recommendation sections mimicking standard clinic workflows.
Contextual Speaker Diarization
Uses GPT-4o to analyze the dialogue context and accurately distinguish between "Doctor" and "Patient" speakers, bypassing the need for specialized multi-channel audio hardware.
Medical Keyword Highlighting
Automatically extracts and color-codes relevant medical terms directly within the transcript UI to help providers scan information quickly: symptoms (red), medications (blue), and diagnoses (purple).
On-Demand Medical Billing Extraction
Analyzes the final SOAP notes to automatically suggest relevant Evaluation & Management (E&M) CPT procedure codes and ICD-10 diagnosis codes to assist medical coding departments.
Interactive Editing Workflow
Features a sleek editing mode that transforms static SOAP notes into input forms, allowing the clinician to easily review, modify, and append the AI's first draft before finalization.
Ephemeral Audio Processing
Processes audio and generates data directly in-memory via Next.js serverless API routes. No database is required, ensuring zero permanent storage of patient audio or PII (Personally Identifiable Information).
Multi-Format Export
Enables one-click copying and downloading of the perfectly formatted transcript, SOAP notes, and billing codes into TXT or Markdown files.
Containerized Architecture
Uses a highly optimized, multi-stage Next.js Docker deployment for seamless scaling and reproducible experimentation within the Cloud2 Labs Innovation Hub environment.