Walkie
Dictation + meetings + read aloud in one app on-device
50 followers
Dictation + meetings + read aloud in one app on-device
50 followers
Walkie turns speech into polished text in any app, gives you meeting transcription with notes and who said what, and reads anything aloud in 950+ voices — free and fully on-device when you want. Mac, Windows, Linux, and iPhone.
How does it work with notes or previous meetings we need to review? Does it provide a brief walkthrough? Is this technically possible if all necessary materials are in the database?
@dorisslane Is the desire to import old meetings into Walkie? If so, we currently have import specifications written for granola and otter specifically. If there's a specific tool that you're using to export your meetings, please submit a support ticket and we can definitely try to get it in there for you.
I didn't understand your last question fully. My mind reframing it?
I dictate a lot, on Vowen right now. What are three reasons I'd switch to Walkie? I think the thread would want them too.
@justin_rockmore great question and Vowen genuinely looks like a great product. I have not tried it myself so I can’t speak to its quality just yet.
From what I read hear are three things that stand out.
Out privacy focus is unparalleled with any other tool in the industry.
Our on-device local tools for dictation, meeting transcription and text to speech are better than any other tool I have tried.
We beat all the completion in price and offer better and more features.
There are definitely more but hose are the top.
@adam_perlis Appreciate the straight answer. Price and privacy I get. What keeps me on Vowen is the corrections it's already learned, and starting that over is what a switch actually costs.
@justin_rockmore You're right that corrections are the real cost of switching, so I built a fix for it just now. In the next update which should land today: Go to Tools → Dictionary → Import from another app → Other. Screenshot your Vowen dictionary, drop the screenshots in, and Walkie reads them right on your computer. "Heard → written" pairs come over as corrections, and single words as words. You review everything before a single word is added.
Amazing product with an amazing founder!
@samraaj_bath1 thank you!
the "who said what" part is the claim I'd want to poke at before trusting it for real meetings. speaker diarization is hard even server-side with a clean single-mic recording, and it gets meaningfully harder on-device with a laptop mic picking up a call through speakers or a crowded room. does it fall back gracefully, just drop the speaker labels and give a flat transcript when it's not confident, or does it guess and assign a speaker anyway? a wrong "who said what" in meeting notes is worse than no attribution at all.
@galdayan your right that is a challenging problem to solve. We handle it a few ways, one diarization can happen locally or in the cloud. The models we work with support both. So once we have who the speakers (Speaker 1 or Speaker 2). Then its figuring out what the name of the person was who said. Generally in a call a person will say another person name and that person will respond. Using the LLM we can process the post call and figure out the most likely speaker. Also we recognize its still a guess, so we also add the ability for our users to see the persons name with a ? beside it in the transcript and they can click it and choose a name mentioned in the call or type one of their own. This will then label all the speakers accordingly. Its not perfect but really no diarization is. Anyone claiming so with the tech we have today is likely lying.
My best thinking happens while I'm pacing around the kitchen, and by the time I sit to type it's gone flat. Talking it out and getting something clean back fits how my head works
@robin_de_lacroix I could not agree more.