One expert hack per tool, every morning.

AI hallucinations and the scattered nature of client notes can undermine the integrity of your health records. Using Whisper locally can help you retain control over sensitive data, allowing you to convert spoken words into organised, reliable documentation without relying on cloud services.

1. Generate a basic transcript — You need a clean text version of your audio before any analysis. This ensures data privacy since no audio leaves your device.

Try it now: From your terminal, run `whisper "your_consultation.m4a" --model tiny.en --output_format txt`. Replace `"your_consultation.m4a"` with your audio file's path.

1. Transcribe specific segments — Not every part of a long recording is relevant for your notes. Focus on key discussions to save time and reduce noise.

Try it now: Add timestamps to your command: `whisper "your_consultation.m4a" --model tiny.en --output_format txt --initial_prompt "The client discussed their morning routine from 10 minutes 30 seconds to 12 minutes 15 seconds."` Adjust the prompt with your relevant time markers.

1. Correct transcription errors — Even the best models make mistakes, especially with medical terminology or accents. Manual review ensures accuracy.

Try it now: Open the `.txt` file generated in step 1 or 2. Search for common misinterpretations of medical terms and correct them directly in the text editor.

1. Extract key themes with a custom prompt — Beyond transcription, you need to pull out actionable insights or recurring patterns from client dialogue. This lays the groundwork for structured note-taking.

Try it now: After transcription, copy the text into your preferred local LLM interface. Use the prompt: "Summarize the key themes from this consultation, focusing on client concerns, reported symptoms, and potential interventions discussed: [PASTE TRANSCRIPT HERE]".

1. Identify client goals — Clearly defining client goals is crucial for treatment planning and measuring progress. Automate this extraction to ensure consistency.

Try it now: With the transcribed text, use a local LLM and the prompt: "From the following transcript, identify and list all explicit and implicit client goals discussed: [PASTE TRANSCRIPT HERE]".

1. Create a structured SOAP note outline — Standardised note formats like SOAP (Subjective, Objective, Assessment, Plan) improve clarity and compliance. Use Whisper's output to populate these sections.

Try it now: Transcribe your session. Then, using a local LLM, prompt: "Convert the following consultation transcript into a SOAP note outline. Identify Subjective (client-reported), Objective (observations/measurements), Assessment (your interpretation), and Plan (next steps): [PASTE TRANSCRIPT HERE]".

1. Flag follow-up actions and reminders — Ensure no action items are missed. Automating the identification of these points helps maintain continuity of care.

Try it now: After transcribing, input the text into your local LLM with the prompt: "Review this transcript and list all agreed-upon follow-up actions for both the client and practitioner, along with any dates or deadlines mentioned: [PASTE TRANSCRIPT HERE]".

This workflow layers directly into your Protocol for client management.