Transcription

Understand how PatientNotes converts your patient encounters into accurate, speaker-labeled transcripts.

Last updated: December 2025

Real-Time Transcription

As you record your patient encounter, PatientNotes transcribes the conversation in real-time. You'll see text appearing on the Transcript tab as you speak, allowing you to verify the recording is working correctly.

Real-time transcript with automatic speaker identification
Processing Time
There may be a 1-2 second delay between speaking and seeing the text appear. This is normal and allows the AI to produce more accurate transcriptions.

Speaker Identification

PatientNotes uses AI-powered speaker diarization to distinguish between different voices in the conversation. Each speaker is labeled to keep the transcript organized.

Automatic Speaker Labels

PatientNotes automatically identifies and labels speakers as "Clinician" and "Patient" based on voice patterns and conversational context.

How Speaker Identification Works

The AI analyzes several factors to identify speakers:

  • Voice characteristics: Pitch, tone, and speech patterns
  • Conversational context: Medical terminology usage, question/answer patterns
  • Spatial cues: Direction and distance from the microphone
  • Learning: The system improves with use as it learns your voice patterns

Multiple Speakers

PatientNotes can handle conversations with more than two speakers, such as when a family member is present or during a consultation with another provider. Additional speakers are labeled as "Speaker 3," "Speaker 4," etc.

Improving Accuracy
The system is most accurate when speakers take turns and don't talk over each other. Brief overlaps are handled well, but extended simultaneous speech may affect accuracy.

Viewing the Transcript

The Transcript tab displays your conversation organized by speaker, with timestamps for reference.

Transcript Features

Search

Use the search function to find specific words or phrases in your transcript. Helpful for long encounters.

Edit Mode

Click on any section to edit the transcript directly. Corrections here will improve the generated note.

Editing the Transcript

While transcription is highly accurate, you may occasionally need to make corrections. Editing the transcript before generating notes can improve the final output.

When to Edit

Consider editing the transcript when:

  • Medical terms are misspelled or incorrect
  • Drug names or dosages are wrong
  • Patient names or identifiers need correction
  • Important details were missed or misheard
  • Speaker labels are incorrect

How to Edit

  1. Switch to the Transcript tab
  2. Click on the section you want to edit
  3. Make your corrections directly in the text
  4. Changes are saved automatically
Focus on Key Information
You don't need to fix every small error. The AI note generation is smart enough to work around minor transcription errors. Focus on correcting critical clinical information.

Transcription Accuracy

PatientNotes uses state-of-the-art speech recognition optimized for medical conversations. Accuracy depends on several factors:

Higher Accuracy

  • ✓ Clear audio quality
  • ✓ Minimal background noise
  • ✓ Standard medical terminology
  • ✓ Speakers take turns
  • ✓ Good microphone placement

May Reduce Accuracy

  • • Background noise
  • • Speakers talking over each other
  • • Heavy accents or dialects
  • • Unusual drug names
  • • Very soft speech

Medical Vocabulary

PatientNotes is trained on medical conversations and recognizes a vast range of medical terminology, including:

  • Drug names and dosages
  • Medical conditions and diagnoses
  • Anatomical terms
  • Procedures and tests
  • Medical abbreviations

The system continuously learns and improves its medical vocabulary based on usage patterns.

Common Transcription Issues

The transcript is missing parts of the conversation

This usually indicates an audio issue. Check that:

  • • Your microphone is working properly
  • • The microphone is positioned to capture all speakers
  • • Browser permissions are granted for microphone access
  • • No other applications are using the microphone
Speaker labels are switched or incorrect

Speaker identification can occasionally mix up speakers, especially at the start of a recording or when speakers have similar voices. You can edit the transcript to correct speaker labels, and the note generation will use your corrections.

Drug names are misspelled

Unusual or brand-name medications may occasionally be misspelled. You can correct these in the transcript before generating notes, or fix them directly in the generated note. Consider spelling out unusual drug names during the conversation for better accuracy.

The transcript shows "[inaudible]"

This appears when the AI couldn't confidently transcribe a segment. Common causes include:

  • • Very quiet speech
  • • Overlapping voices
  • • Loud background noise at that moment

If you remember what was said, you can edit the transcript to fill in the content.

Next Steps

With your transcript ready, it's time to generate a clinical note. Learn how to choose templates and create documentation.

Was this article helpful?