Audio Transcription

Convert audio recordings to text using OpenAI's Whisper model, running entirely on your local machine.

Features

• Local Processing: All transcription happens on your device - no cloud uploads
• Multiple Languages: Supports 99+ languages
• Timestamps: Get word-level timing for each segment
• Translation: Optionally translate to English

Usage

1. Open Command Palette → "Audio: Transcribe"
2. Select your audio file (WAV format recommended)
3. Wait for transcription (progress shown)
4. Save as a new note or insert into current note

Supported Formats

For best results, use WAV format. Other formats require ffmpeg to be installed.

Model Selection

Whisper comes in different model sizes:

Default is base.en (English-optimized base model).

First-Time Setup

On first use, Naidis will download the Whisper model:

1. Settings → AI → Download Whisper Model
2. Select model size (base recommended)
3. Wait for download (~150 MB for base)
4. Model is cached locally for future use

Use Cases

Meeting Notes

Record your meetings and transcribe them to searchable notes.

Voice Memos

Quick voice notes become markdown files.

Podcast Notes

Transcribe podcast episodes for reference.

Interview Transcription

Convert interview recordings to text.

Output Format

Transcriptions include:

# Audio Transcription

**Duration**: 5:32
**Language**: English

## Transcript

[00:00] Hello and welcome to today's meeting.
[00:05] We have several items on the agenda.
[00:12] First, let's discuss the project timeline.
...

Tips

• Quiet environment: Better audio = better transcription
• WAV format: Use WAV for fastest processing
• Shorter clips: Split long recordings for better results
• English model: Use .en models for English-only content (faster)

Requirements

• macOS, Windows, or Linux
• ~500 MB disk space for model
• 4+ GB RAM recommended