Documentation

Learn how to use Notably

Transcription

AI-powered speech-to-text with two processing modes for flexibility and accuracy

Overview

Notably offers two transcription modes to fit different workflows. Real-time transcription shows text as you speak during recording, while post-recording transcription uses more powerful models for higher accuracy after recording ends.

All transcription processing happens locally on your device using Whisper AI models. No audio data is ever sent to the cloud.

Transcription Modes

Real-time Transcription

Live transcription during recording using lightweight Whisper models. Text appears as you speak, providing immediate feedback.

  • Fast processing: Uses optimized lightweight models (Tiny, Base, Small)
  • Live feedback: See transcription text as recording progresses
  • Lower latency: Minimal delay between speech and text display
  • Good accuracy: Suitable for most general meetings and notes

Post-recording Transcription

High-accuracy transcription after recording completes using more powerful models. Can be automatic or triggered manually.

  • Higher accuracy: Uses larger models (Medium, Large) for better results
  • Automatic or manual: Configure to run automatically or generate on demand
  • Better punctuation: More accurate capitalization and sentence structure
  • Technical terms: Improved recognition of specialized vocabulary

Language Support

Notably supports automatic language detection and 13 specific languages for transcription:

English
Spanish (Español)
French (Français)
German
Italian
Portuguese (Português)
Dutch
Polish
Russian
Japanese
Chinese
Korean
Arabic
Hindi

Use auto-detect to let Notably identify the language automatically, or select a specific language for improved accuracy.

Transcript Viewer

View and interact with transcriptions using an advanced segment-based display:

  • Segment display: Transcriptions organized into logical segments with timestamps
  • Word highlighting: Individual words highlight as they're spoken during playback
  • Jump to segment: Click any segment to jump to that point in the recording
  • Sync to playback: Transcript scrolls automatically as recording plays

Re-transcribe Recordings

Generate new transcriptions anytime with different models or settings:

  • Switch to a larger model for improved accuracy
  • Try a different language setting if auto-detect was incorrect
  • Re-process after downloading a new model version
  • Keep multiple transcription versions for comparison

Original recordings remain unchanged - re-transcription creates a new transcript without affecting existing ones.

Privacy & Local Processing

All transcription processing happens entirely on your device:

  • Whisper models downloaded and stored locally
  • Audio never leaves your device
  • Transcription text stored in local database
  • No internet connection required after model download
  • Complete data ownership and control

Model Selection Guide

Tiny / Base

Best for real-time transcription. Fast, lightweight, good for general meetings.

Small

Balanced option. Better accuracy than Tiny/Base while still fast enough for real-time use.

Medium

High accuracy for post-recording. Good for important meetings, technical discussions.

Large

Highest accuracy. Best for critical content, difficult audio, or specialized vocabulary.