Feature Requests

One of my favorite features about Better Dictation is that when I start to record, it lowers the volume on everything else being played.

Hi, I don't know if this is technically feasible and how complex, but it will be great. It's very niche, but it's a super nice detail. I usually work with music. When I am about to speak to superwhisper, I need to pause the music, talk, and then resume the music. If I have the music while I talk I get distracted. It would be awesome if superwhisper can pause any music for Spotify and other music apps (or any source) when I start talking. And then, resume it again.

For shorter dictations, such as brief replies to people or ChatGPT, I sometimes find it inconvenient to press my key command twice. This can create resistance in using Superwhisper compared to simply typing. It would be great if holding down the shortcut could activate a 'push-to-talk' mode. In this mode, Superwhisper would detect the keypress and initiate dictation. After approximately 2 seconds, if the keys are still held down, the mode would automatically engage. Releasing the shortcut would then stop the transcription and paste the text.

Add Llama 3 via Groq as an available API provider for its fast response time to improve Pro user latency. https://console.groq.com/docs/quickstart

The only local LLM models I can use in this app are specific for the English language, but I speak in French. I would like to be able to use another LLM, or my local Ollama instance, to avoid having to be online to benefit from the language model pass.

My main use case for superwhisper is to transcribe meetings using window recording. While the output is already very useful, it would be even better if the result would be split by speaker. This could be a per-profile configuration, allowing separating a "dictation" use case from a "meeting transcript" use case. Optional follow-up idea: Allow identifying speakers at timestamps while the recording is ongoing to already tag the result with appropriate speaker names in the output.

with the ability to press the new side button to access

I use the function that it doesn't automatically insert the text. And here I would like there to be a short acoustic signal as soon as the text is ready, so that I know immediately, without looking at the icon in the menu bar, that it is finished and I can insert it immediately while I still remember. Otherwise, there is a risk that I will completely forget it. I have seen that there is an option in the settings called "Sound Effects Enabled". I have activated this, but unfortunately, I still don't hear any sound.

It would be extremely useful if I could use superwhisper to transcribe sources of data other than immediate recordings such as my voice notes from my phone, or other recordings I have around. I'm not exactly sure about the user interface here, but it would be useful if I could just drag in a bunch of files, whether into the UI or even into a subdirectory inside the recordings folder, which would be auto-detected and transcribed and given the same kind of text file as a companion output.

90% of the time that I use Superwhisper, I'm dictating into ChatGPT or bard, I press enter, and wait for a response from the LLM. My workflow would be even faster if Superwhisper handled the speech-to-text-to-LLM process and provided a response directly.