what is lazymeet fr?

LazyMeet is a desktop app that turns meeting recordings into transcripts and summaries entirely on your own computer, with no cloud service, no accounts, and no API keys needed. You drag a recording, such as a screen capture from OBS, onto the app window, and it produces a transcript broken down by speaker, along with a summary and a list of action items. Under the hood the app is built as a single Rust binary using the Tauri framework. When you add a recording, it converts the audio, detects when people are speaking, runs speech to text on those sections, figures out who spoke when, and then feeds the transcript to a small local language model to generate the summary and takeaways. All of these AI models download once the first time you run the app, a process of a few gigabytes, and after that everything works fully offline. A notable feature is speaker recognition across recordings. Normally each new recording numbers its speakers starting over, so Speaker 1 in one meeting is not the same person as Speaker 1 in another. LazyMeet lets you rename a speaker once, and it saves a voice fingerprint for that person in a local file. On future recordings, it tries to match detected voices against this saved list and suggests a name, which you can confirm or reject. Confirming teaches the system more about that person's voice over time, while rejecting simply undoes the guess without corrupting what it already learned. Before transcribing, you get a preparation screen where you can preview the recording's waveform and choose to skip parts you do not want transcribed, plus set the expected number of speakers. On a real 28 minute test recording on an Apple laptop, the whole process took a little over 3 minutes, roughly nine times faster than the recording's own length. Nothing about your audio, transcripts, or voice data ever leaves the machine, and you can configure the app to delete each recording's audio automatically once processing finishes. Building it from source requires Rust and a few native libraries. It is released under the GPL v3 license.