The `Ultra` voice model definitely has some flaws with language detection. Very unbearable. Because of that, I now got every mode I create in German and in English because I need both, which is super inconvenient. Also, on macOS and iOS I got an English translation even if I don't use any prompting model and set it explicitly to German. This only appeared like a few weeks ago. It was much better before that.
No matter how this feature request would end up, I would at least strongly want to keep the option that selecting an input language and eventually in a custom prompt defining a different output language remains a valid configuration. It's a use case that I have very often
this works in the Mac OS version, but not in the iOS version...
Yes, just struggled with that this morning, as the app I was using, was shifting modes and I had a mode for German language, but it would consistently transcribe in English. Eventually solved it with a prompt, but took me forever to find out what was going on. Basically, app context overrides mode (so if I wanted to use two different language modes and custom prompts in the same app, I currently can't do that).
This seems like a good and, at the moment, the only solution, but it is then processing the voice input through the voice model into English (or any other unwanted language) and then back through your pre-prompting model into your desired language. It's definitely a bit slower, and in some cases loosing `meaning` on the way. Not ideal.
yes, please
Hey, I’m also curious which/what prompt you used! From what I’ve seen, auto transcription usually sticks to whatever settings is set in Superwhisper. I’d love to hear how you handle this, because I swap between two languages a lot too and having to keep switching profiles/modes is a bit of a pain.
What prompt did you use?
You need to scroll until languages that start with "A" and there will be Automatic mode, which is basically what you are looking for. Works good for me, easily switches between English and Russian.