| I'd like to create a voice recording app for Ubuntu Linux. The app should have the following functionalities. It's a voice recorder, and it has the essential voice recording functionalities of record, pause, stop, and restart. The restart scraps the current recording in cache and restarts the recorder from zero.<br><br>For the transcription process, I'd like to institute the following workflow. We'll use Google Gemini API and ensure that we're using Gemini 2.5, which supports multimodal input, including audio. The recording captured from the user should be optimized for this purpose of voice capture for speech to text. By which I mean, I would suggest that we record in mono. We capture the recording in a space-efficient format. We're optimizing for creating a voice recording that is not necessarily the greatest and most detailed of audio clarity, but which strikes the best balance between quality and space efficiency for transcription.<br><br>The voice recording will get sent to Gemini for transcription with a system prompt that instructs it to transcribe and also clean up the recording by removing filler words, adding sentence structure, and adding spaces. There can be a second button which is called transcribe and optimize, and the transcribe and optimize workflow is the same except that the system prompt is a little bit more instructive and it tells Gemini, in addition to those steps, to remove filler words, add sentence structure, paragraph spacing, and try to optimize the text by adding headings, organizing the thoughts a bit, and removing repetition, so it's a little bit more aggressive.<br><br>In both cases, the transcription, when it returns from Gemini API, will populate into a... In fact, Gemini should return two things, a title and a text. The title is a short title for the voice note. The text is short is the text, and so on. The text is formatted in Markdown; it should appear within the Markdown within the text editor. There should also be a clipboard button, and finally, the user should be able to save the note.<br><br>When the user chooses to save the note, it will get saved to a predetermined folder which the user selects as where they save voice notes on their operating system. And it's saved there as a Markdown file with the title in machine-friendly format. So if the voice note title has spaces, the saved file name will just replace those with hyphens.<br><br>The app would be run repetitively such that if the user wants to record a new note, they start again, and when they do transcribe or transcribe and optimize, it will send and then overwrite the previous transcription. So the user has to click the save button in order to save it, or there can be an option for auto-saving configurable as a user setting. |