Semantic search
You can now search your meeting history by meaning instead of exact words. It all runs on-device, and it works across languages, so a Japanese query can find an English meeting.
Some notes on how this got built, because it didn't go the way I planned. My first version used Apple's built-in embedding API (NLContextualEmbedding), mainly because it needs no model download. I really wanted that to work. But it's a general-purpose model, not tuned for retrieval: short filler lines like "yeah, okay" matched almost any query, and the similarity scores moved around so much that I couldn't find a threshold separating real hits from noise. I tried the usual tricks (centering the vector space, weighting by length) and they helped, but on my real transcripts it was still a coin flip. So I gave up on it and switched to Qwen3-Embedding-0.6B, a model actually trained for retrieval, running locally through llama.cpp. That costs users a one-time 640 MB download, which I was reluctant about, but the results on my own messy meeting data were good enough that I stopped second-guessing it.
Foyer
recording the demo with wifi off is the whole pitch in one move. on-device is the only trust model that makes sense once real business talk is in the room
QuickQuill
@fberrez1
Thank you! I always believed showing a demo beats making claims. Running it with Wi-Fi off was my way of proving nothing gets sent to a server. When important business is being discussed, you really don't want a cloud service in the room.
Runs offline and still pulls off live translation mid-meeting, which genuinely surprised me. The no-account, one-time-purchase setup is a nice change from the usual subscription grind.
QuickQuill
@erafettinrqqxz
Thank you! It really is surprising that it keeps working even with the network off. And when we're already juggling so many subscriptions, the last thing anyone wants is to add another one. That's why I made it a one-time purchase!
@Taisei that's a really clean answer, thanks. makes sense that spooling raw audio is basically free and the real work happens after you stop recording. good to know live subtitles ride the same on-device stack too rather than needing something heavier running the whole time
runs fully offline and still nails live translation while recording, pretty rare combo for a mac app. the no bot joining the call thing is a nice touch too.
QuickQuill
@hirabakrdvnljo
Thank you! I actually built a cloud-based voice AI agent before, and I never imagined that recording, transcription, summaries, and live translation could all be done locally. And that's what makes the no-bot approach possible. Everything stays local :)
Taisei, keeping private conversations on my own machine is exactly the kind of thing I have wished for. Trust matters a lot to me here, and knowing nothing quietly wanders off elsewhere makes me far more relaxed about it.
QuickQuill
@benoit_rq
Thank you, that's exactly the feeling I wanted to build for. Once your conversations stay on your own machine, you stop having to think about where they might end up. That peace of mind is the whole product, really.