Speakey vs MacWhisper

At a glance

Name	Best for	Price	File transcription	Voice commands
Speakey	Dictation and Mac control	FreePaid onceFree plan; Pro $4.99 once	No	Yes (Agent)
MacWhisper	Files, meetings, subtitles	FreePaid onceFree; Pro €64 once	Yes	No

Two different jobs

MacWhisper started as a way to drop an audio file onto a window and get a transcript back, and that is still where it is strongest. Its feature list covers mp3, wav, m4a, ogg, opus, mov and mp4 files, YouTube and media URLs, batch jobs, automatic speaker recognition, podcast transcripts split by speaker, .srt and .vtt subtitles, transcript translation and meeting recording for Zoom, Teams, Webex, Discord and others. It also has system-wide dictation, but that is one feature among many.

Speakey has no file transcription. It is a hold-to-talk layer: you hold a key, speak, and the text is typed where your cursor is. Around that it adds Rewrite (your words passed through a Rule before typing) and Agent (spoken commands that open apps, place windows, read your calendar, draft Gmail replies or work through a web form).

Price

MacWhisper's site lists a free version ("Free forever") and MacWhisper Pro at €64 per license, paid once, with lifetime updates. It is sold on Gumroad and through the Mac App Store; the site mentions differences for the App Store version but does not spell them out on the main page. The site does not list how many Macs one license covers or which features are Pro-only, so check the FAQ before buying.

Speakey is free with weekly limits (2,000 words across its modes plus 5 minutes of Live Listen, unlimited for the first 3 days). Pro is $4.99 once for up to 3 Macs, an early adopter price that is normally $9.99, with a 7-day self-serve refund. No account needed.

Engines and privacy

MacWhisper runs Whisper locally and mentions Parakeet and other models. It also lets you bring your own API keys for OpenAI, Anthropic, Groq, Google, OpenRouter and others, or use local language models through Ollama and LM Studio, for cleanup and prompts. It lists 100 languages.

Speakey runs four engines on the Mac: Parakeet TDT 0.6B v3 for 25 European languages, Qwen3-ASR for about 30 languages including Indonesian (macOS 15 or later), Whisper Large v3 Turbo through WhisperKit for about 99 languages, and Apple Speech. It picks one based on your Mac and the languages you speak. Audio never leaves the Mac. Rewrite and Agent send text to your own Claude, ChatGPT or Gemini account.

Speed where it matters for each

For file work, what counts is throughput: how many minutes of audio per minute of waiting. For dictation, what counts is the gap between letting go of the key and seeing text. Speakey is tuned for the second. On an M5 Max in Speakey's own tests, Parakeet returned a finished English take in 40 ms at the median, and Qwen3-ASR returned a mixed Indonesian and English take in 30 ms at the median when it could reuse its live preview. Those are Speakey's numbers on one Mac, not a comparison with MacWhisper, which we haven't measured.

Where MacWhisper wins

• Transcribing existing files, URLs and batches.
• Speaker recognition and per-speaker podcast transcripts.
• Subtitles (.srt, .vtt) with an inline player and translation.
• Meeting recording that starts and stops on its own.
• Integrations: watched folders, Notion, Zapier, Obsidian, n8n, Make.com, webhooks and a CLI.

Where Speakey wins

• Live dictation into any app as the main job, with a live preview while you speak.
• Agent mode: voice commands that act on the Mac, not just text.
• Price: $4.99 once for 3 Macs, plus a free plan.
• An engine picked for your languages, including fast Indonesian and English on Qwen3-ASR.

Questions