Why You Need This Skill
If you have ever worked in a hospital or medical lab, you know that there is a massive amount of paperwork to handle every day. Medical records, ward round notes, surgical reports, research papers, theses... If you had to type everything out manually, you would go insane. During my internship, I saw an attending physician who would record voice memos on his phone while doing rounds, then transcribe them at home at night. He often worked until 2 AM. Later, he told me he had tried various speech-to-text software, but none were professional enough—they either failed to recognize drug names like "amlodipine" or misheard "white blood cells" as "white bood sells".
So when I came across the medical-scribe-dictation skill, my first thought was "finally, a solution." It's not just a simple speech-to-text tool; it's a smart assistant specifically optimized for medical scenarios. Besides basic dictation, it has an embedded academic writing module that helps standardize terminology, adjust formatting, and even match reference formats.
What Can It Actually Do?
Simply put, you speak into a microphone: "A 58-year-old male patient presented with chest pain for 2 hours. ECG shows ST-segment elevation, diagnosed with acute anterior myocardial infarction. Immediately administered aspirin 300 mg chewed, clopidogrel 600 mg loading dose, and scheduled for emergency PCI." The skill then automatically outputs a well-formatted, terminologically accurate text. And if you are using it for research dictation—for example, "We conducted a double-blind randomized controlled trial involving 120 patients with type 2 diabetes"—it can automatically expand "double-blind randomized controlled trial" to "double-blind randomized controlled trial (DB-RCT)" and provide reference format suggestions.
Installation Steps Are Not Difficult
Let's start with installation. This skill is based on the medical-research-skills package. Before installation, ensure your computer has Node.js and npm installed (if not, go to the official website to download and install them—it takes a few minutes). Then open your terminal (Mac: Terminal; Windows: Command Prompt or PowerShell) and enter the following command:
npx skills add https://github.com/aipoch/medical-research-skills.git --skill medical-scribe-dictationOnce the download and installation are complete, you will see a "medical-scribe-dictation" directory added to your folder. Inside, you'll find some useful files: SKILL.md is the documentation file; the scripts folder contains audio processing scripts; requirements.txt lists Python dependencies—since the speech recognition backend may use models like Whisper, a Python environment is required. To install the Python dependencies, simply run:
cd medical-scribe-dictation
pip install -r requirements.txtA Few Tips for Usage
- Speak clearly: Although the model supports multiple accents, terms like "left radial artery" should be spoken in standard Mandarin or English. You can first record a test audio to check recognition accuracy.
- Leverage context: The skill allows you to import a medical dictionary or custom terminology list in advance. If your department has special terms (e.g., "TACE procedure" or "ECMO"), add them to the dictionary—this significantly improves recognition precision.
- Don't neglect formatting: The output text defaults to Markdown format, with clear headings, lists, and numbering. If you want to copy directly into Word, you can configure it to plain text or rich text format.
- Keep the recording environment quiet: Don't use it next to an MRI machine... don't ask how I know.
Practical Application Scenarios
I tested it in two real-world scenarios: first, helping a senior colleague organize her outpatient clinic dictation recordings (she sees over 80 patients a day, and her voice was worn out); second, writing the "Methods" section of my own paper. The results were impressive: the recognition accuracy for outpatient records was above 95%, with only a few drug names needing manual correction. When writing the paper, I dictated: "We used SPSS 26.0 for statistical analysis, with continuous data expressed as mean ± standard deviation." The skill automatically added "(mean ± SD)" as an English annotation and even prompted a citation reference (though it requires connecting to PubMed).
For surgeons who frequently need to produce operative notes, this skill is truly practical. You can speak the procedure into the microphone right after surgery, and within a minute, a well-structured note is generated. No more staying up late that night to transcribe from a surgical video.
A Few Pitfalls to Note
One thing to remind you: this skill primarily supports English, but it also has some support for Chinese (depending on the underlying model). If you mainly work in a Chinese-language environment, it's recommended to first check the scripts folder for a Chinese language model configuration file, or download a Chinese speech recognition model yourself and replace it. Also, be aware of privacy issues. If patient data is involved, it's best to run it locally offline and not upload to the cloud. The skill supports local inference by default—just place the model files locally.
Final Thoughts
Overall, the medical-scribe-dictation skill gives me the feeling that someone has finally taken the time to create a smart assistant for the medical field. It's not just a transcription tool; it bridges the gap between dictation and writing. For healthcare professionals who deal with text every day, it's definitely worth trying. But remember, no matter how good the tool, it's only an aid—you still need to double-check the key information yourself.
Well, that's all I have to share for now...
