Back to Blog
v1.5.0

Pindrop v1.5.0: Multi-Engine Support with Parakeet ASR

Major release introducing pluggable transcription engines, Parakeet integration via FluidAudio SDK, and enhanced audio capabilities including VAD, diarization, and streaming.

We’re excited to announce Pindrop v1.5.0, a significant evolution in our quest to provide the best offline dictation experience for macOS. This release introduces a major architectural change: support for multiple transcription engines, starting with integration of the high-performance Parakeet ASR models.

Multi-Engine Architecture

The biggest change in v1.5.0 is under the hood. We’ve restructured the transcription system around a new TranscriptionEngine protocol, allowing Pindrop to support multiple ASR backends:

  • WhisperKitEngine — Our battle-tested WhisperKit integration (default)
  • ParakeetEngine — New high-performance option using FluidAudio SDK

This modular design means we can now add support for additional engines without rewriting the entire app. You can switch engines at runtime from Settings → Models, and each engine can expose its own capabilities and model options.

Parakeet ASR via FluidAudio

We’re now shipping with support for Parakeet ASR models through the FluidAudio SDK:

  • Parakeet V2 — Optimized for English, ~2x faster than Whisper while maintaining accuracy
  • Parakeet V3 — Multilingual support coming soon

Parakeet models are particularly impressive on Apple Silicon, utilizing the Neural Engine for real-time transcription with minimal CPU impact. To try it:

  1. Open Settings → Models
  2. Download the Parakeet V2 model (~150MB)
  3. Select it as your active model
  4. The engine switches automatically

Extended Audio Capabilities

We’ve laid the groundwork for advanced audio features with new protocol definitions:

  • Voice Activity Detection (VAD) — Automatically pause recording during silence
  • Speaker Diarization — Identify different speakers in transcripts (coming in v1.6)
  • Streaming Transcription — Real-time word-by-word output (Parakeet only)
  • Text-to-Speech — Foundation for future voice feedback features

These capabilities are exposed through the engine system, so each ASR backend can advertise what it supports. WhisperKit provides VAD and language detection; Parakeet adds streaming and will support diarization in upcoming updates.

Feature Model Downloads

Optional “feature models” can now be downloaded separately from transcription models:

  • VAD models for silence detection
  • Diarization models for speaker identification
  • Streaming enhancement models

Enable features in Settings → Feature Models, and Pindrop downloads the required components on first use.

UI Polish

Small details matter:

  • SF Symbols throughout the status bar menu for better visual scanning
  • Active model indicator — Green dot shows which model is currently loaded
  • Provider badges in the model picker (WhisperKit, Parakeet, etc.)
  • Cleaner Settings menu organization

What’s Next

We’re actively working on OTA Updates — automatic in-app updates so you always have the latest improvements without manually downloading releases.

Upgrade Notes

v1.5.0 is fully backward compatible. Your existing WhisperKit models continue to work exactly as before. If you want to try Parakeet, simply download it from Settings.

Download: GitHub Releases

As always, feedback is welcome. Happy dictating!