Feature Requests

Specifically it stopped pasting the result. Result is transcribed correctly, and is available in history, but auto-pasting no longer works. Happened after I restarted my Mac for unrelated reasons, but now I can't get it to work. I guess I'll have to go look at alternatives, typing by hand like a caveman is no longer okay.

I said "I need the instruction itself, not the implementation. Please format everything in Markdown.". I expected this to be transcribed to text as normal, but instead the response was an instruction to how to use Markdown in prompts. It seems that this message escaped the system prompt and was interpreted as a direct instruction to the post-processing LLM (Claude 3.7 Sonnet).

Since the last update, the Ultra Cloud Model sometimes adds double spaces when I do a pause and speaking. It seems to happen more in pure voice mode than in more language-processed modes. The paragraph above was dictated and you can see that there is a double space between the two sentences.

When I press a push-to-talk button, but say nothing, I get a weird output all the time. The output contains "Closed Captioning by ". For example: "Closed Captioning by blog.com " "Closed Captioning by btistudios.com " "Closed Captioning by www.closedcaptioning.com " My voice model is Ultra, my language model is Sonnet 3.5

Since the last update, the Ultra Cloud Model sometimes adds double spaces when I do a pause and speaking. It seems to happen more in pure voice mode than in more language-processed modes. The paragraph above was dictated and you can see that there is a double space between the two sentences.

I like the combination space + fn to toggle the recording. However, when I try to set it up, it only registers the fn key and space briefly, then it removes the space key from the setup.

I use a 192khz audio input on my computer and Superwhisper seems to be unable to receive input from this device.

Even when I'm not recording anything and it's not active, I'm seeing that it consumes close to 50% of the CPU.

When I'm using a full screen app, the recording window is not shown at all, instead it is shown on the desktop instead of the full screen app I'm currently using. Can this be fixed so that the recording window floats on top whatever app you're in, like in the MurmurType app?

open "superwhisper://mode?key=message-ONQC" open "superwhisper://record" --- When launching Superwhisper via deep links like those above, the input field loses focus, requiring a manual click to refocus. It would be greatly appreciated if this could be improved so that the input field retains focus when launched through deep links.