Linked List: November 25, 2023

Audio Hijack 4.3’s New Transcribe Block 

Rogue Amoeba:

Transcribe can convert speech from an astonishing 57 languages into text, providing you with a written transcript of any spoken audio. It’s powered by OpenAI’s automatic speech recognition system Whisper, and features two powerful models for fast and accurate transcriptions.

Best of all, unlike traditional transcription services, Transcribe works for free inside of Audio Hijack. There’s absolutely no on-going cost, so you can generate unlimited transcriptions and never again pay a per-minute charge. It’s pretty incredible.

It’s also completely private. When you use Transcribe, everything happens right on your Mac. That means your data is never sent to the cloud, nor shared with anyone else.

This makes for a perfect one-two shot with Retrobatch 2: Audio Hijack is also a node-based media tool (which predates Retrobatch), and this new Transcribe block is also putting powerful machine learning tools into an easily accessible form.

This Transcribe feature in Audio Hijack is also an exemplar of the power of Apple silicon — it works on Intel-based Macs too, but it’s just incredibly fast on Apple silicon (I suspect because of the Neural Engine on every M-series chip).

Retrobatch 2.0 

Gus Mueller, writing at the Flying Meat blog:

In case you’re not aware, Retrobatch is a node-based batch image processor, which means you can mix, match, and combine different operations together to make the perfect workflow. It’s kind of neat. And version 2 is even neater. [...]

Retrobatch is obviously not Flying Meat’s most important app (Acorn would fill that role), but I really do like working on it and there’s a bunch more ideas that I want to implement. I feel like Retrobatch is an app that the Mac needs, and it makes me incredibly happy to read all the nice letters I get from folks when they figure out how to use it in their daily work.

Five years after Retrobatch 1 shipped, I’m happy to see version 2 out in the world. And I can’t wait to see what folks are going to do with it.

“Node-based batch image processor” means that you design and tweak your own image processing workflows not with code, but through a visual drag-and-drop interface. (But you can use code, via nodes for JavaScript, AppleScript, and shell scripts.) You can program your own highly customized image processing workflows without knowing anything about writing code. It’s useful for creating workflows that work on just one image at a time, but Retrobatch really shines for batch processing.

There are a zillion new features in version 2, but the star of the show has to be the new “ML Super Resolution” 4× upscaler: a powerful machine learning model made easily accessible.

Teenage Engineering’s EP–133 K.O. II Sampler 

I can’t read or play music, and struggle even to clap to a beat, so I would have zero use for this device. But I still want to buy one. Just look at it. Absolutely gorgeous.