Audio specialist Rogue Amoeba has added a useful new feature to its recording tool Audio Hijack Plugged in: With version 4.3, audio content can also be transcribed automatically. There is also a new “Transcribe Block” that you can insert into your recording setup using a template or simply drag & drop. Whisper serves as the technical basis Open source based transcription engine by OpenAI.
Advertisement
Open source ASR directly on recording
Whisper is a so-called ASR system (Automatic Speech Recognition). It has been trained with around 700,000 hours of language in various tongue strokes, so it understands not only English, but also German and other languages. The system is considered robust and accurate; To achieve the quality offered, you previously had to use commercial services. There have been various free and paid options for a long time Mac clients for Whisper in the App Store and elsewhere. Integrating a transcription into a recorder app makes it even more convenient to use.
Audio Hijack records audio from a variety of sources on the Mac – for example, only from pre-selected apps, audio inputs or devices. The app contains numerous filters and setting options for routing audio. Audio Hijack has been available for many years and is popular among podcasters, among others. There was a major rewrite last year with version 4.
No additional costs, but still beta
Using Whisper has the advantage that the transcription takes place directly on the Mac and not in the cloud. As is usual with OpenAI API access, there are no additional costs; you can have as much transcribed as you want. The Transcribe Block can be inserted directly behind an audio source; this can also be the microphone to dictate texts. It is also possible to transcribe audio calls or YouTube videos from a browser. Users can choose between different Whisper modes, including the very precise one. The transcript is time stamped.
Transcription in Audio Hijack 4.3 is currently still in a beta phase (which does not apply to the rest of the app). If problems are discovered, you should report them to Rogue Amoeba. Audio Hijack is not marketed as a subscription, but as a one-off payment. This currently costs $77 including German sales tax. Audio Hijack 3 upgrades cost $35 plus tax.
(bsc)