MacWhisper 8 debuts with new video player, WhisperKit models, ChatGPT 4o support

Zac Hall | May 20 2024 - 8:30 am PT

MacWhisper is a super useful utility for quickly transcribing spoken audio on the Mac. Version 8 is launching today with new features around video, speed, AI, and more.

Video

A new video player is the flagship feature of MacWhisper 8. This greatly improves the experience when transcribing audio from video files.

Transcribed audio is displayed as subtitles, and translated text appears as separate subtitles. The video player appears inline by default, or you can optionally turn it into a separate window.

Speaking of video, MacWhisper 8 also delivers improvements around YouTube. Downloaded video can play in the mini-player, downloads can be high or low quality, and YouTube downloads are generally just faster. You can also choose between video or audio-only downloads from YouTube.

WhisperKit

MacWhisper 8 also introduces WhisperKit model support.

WhisperKit is a Whisper engine that delivers real-time, on-device speech recognition on Apple silicon. WhisperKit models take advantage of hardware acceleration including the Neural Engine and Apple’s CoreML framework, making it especially efficient.

WhisperKit models can be added to the model list in Settings > Advanced > Show WhisperKit Models.

AI and improvements

Lastly, MacWhisper 8 introduces support for the latest ChatGPT models from OpenAI. This includes GPT-4 Turbo and the newly launched GPT-4o that premiered a week ago.

Meanwhile, MacWhisper includes a variety of general improvements to the app that you can read below:

If you have a character limit set, the app will not cut off words in the middle of a word.
New menubar icon that doesn’t conflict with the standard microphone icon.
Quality and language selectors moved to the toolbar. Expand your window if they’re not visible.
Opening .whisper files is now possible while models load.
Updated to the latest Whisper C++ engine, now with Flash Attention (activate in Settings > Advanced).
Redesigned Manage Models screen for easier model selection. Feedback is welcome.
Enhanced error handling for model downloads.
“MS Teams Virtual Mic” excluded from microphone options as it’s not an actual mic.
Fixed a bug where invalid license error codes weren’t displayed.
Resolved a crash when non-pro users added more than two speakers.
The Esc key won’t close screens during active processes like recording or batch transcription.
The Cloud Transcription feature now only lists the languages that are supported (57) compared to the 100 that are supported locally
Fixed the bug where m4a/mp4 files were being rejected even though they are supported
All the file formats that the local transcription mode supports are now supported for Cloud as well
Keyboard shortcut modifiers now displayed in the UI (⌘+R etc).
Improved button design.
Fixed transcript copy errors.

MacWhisper is a core app on my Mac and a must-have if you’re AI curious. Grab MacWhisper 8 today and be amazed at the power of AI transcription on your Mac.