Hi there,
Could anyone maybe explain the Whisper Memos privacy policy in plain English? They refer largely to OpenAI, but what does that effectively boil down to? Would Just Press Record be a more private solution?
@MacSparky, might I ask you to share your (non binding 
) thoughts on this? Either during a show or in writing here maybe?
Have a nice day everybody!
Tom
I think Just Press Record uses Siri dictation. But I donât remember where I got that information, so take it with a grain of salt.
When OpenAI ran out of training data from the web, they developed Whisper, so they could scrape audio, turn it into text and continue feeding the beast. Even if their privacy policy was written at the third grade level, one would have to also be functioning at the third grade level to trust them with anything sensitive. 
You can run it locally, but again, if youâre concerned with privacy, Iâd lock it down and make sure it canât phone home, now or later.
This is my understanding ⌠Whisper definitely works in the cloud. They have a setting where they will delete the file after transcription, but at that point, your audio has already been processed by the transcription machine. I wouldnât use any non-local transcription for private data.
It might also be worth waiting until WWDC before deciding on any AI solution. I would not be surprised if Apple has its own on-device voice transcription native in the OS this year.
OpenAI privacy was not really clear about a year ago. For example, using ChatGPT history meant that you were also agreeing the usage of your data for training their public models (subtle trick, both things are not necessarily related).
In case of doubt, I would suggest use MacWhisper or Buzz, that is, an application that downloads the necessary models from a public repository from Huggingface and then the processing runs locally. I have had success sending regular voice memos from Apple stock app to my Mac via Airdrop and processing the audio in the computer, no the phone.
Note to add, I checked Whisper Memos and it is still not clear to me where the audio transcription processing is done, and then there are also additional privacy concerning features like âwe send the transcription to your emailâ and âGoogle Firebaseâ.
Thanks to all of you who have replied. I have decided to stick with superwhisper; which is part of Setapp on MacOS, uses a local model and optional cloud integration. Sadly no Apple Watch support, let alone the Drafts integration as explained by MacSparky
I use MacWhisper which sounds very similar. I get round the lack of watch support by dictating journal entries into Voice Memos and having MacWhisper transcribe them later. This works well for me.