MacParakeet

MacParakeet is a speech-to-text app for people who want to dictate across Mac apps, transcribe recordings and capture meetings while keeping speech recognition on their own hardware.

• System-wide dictation: Insert spoken text into apps with a text cursor, with custom hotkeys and vocabulary for names or specialist terms.
• Files and media links: Transcribe audio and video files in batches, or fetch audio from YouTube, Loom, Vimeo and podcast links for local transcription.
• Meeting capture: Record system audio and your microphone together, follow a live transcript and take notes during the call.
• Transcript editing and export: Correct saved transcripts and AI results, retain speaker labels when the speech engine supports them, and export formats including Markdown, SRT, DOCX and PDF.
• Optional AI: Generate summaries, ask questions about transcripts, clean up dictation or rewrite selected text using a configured local or cloud provider.

The app is free and open source under GPL-3.0. It requires Apple Silicon and macOS 14.2 or later, with no account required. After the initial model download, dictation and local file transcription work offline. Media links need internet access to fetch audio.

Parakeet uses the Apple Neural Engine and handles English and supported European languages. WhisperKit adds languages including Korean, Japanese, Chinese, Hindi and Arabic. Cohere Transcribe is a local batch option that produces plain text without live previews, word timestamps or speaker labels.

Ollama and LM Studio can keep optional AI processing on-device. Cloud providers such as OpenAI and Anthropic receive the text context needed for the chosen feature, never captured audio. Apple Intelligence supports dictation cleanup on eligible Macs running macOS 26 or later; summaries, chat and text rewriting require another provider. Optional anonymous telemetry excludes audio and transcript content.