Audio Transpile / Speech-to-Text / OpenAI Whisper Transcription Integration
OpenAI's Whisper model does a great job at taking voice recordings and transcribing them perfectly. This can be great for transcribing meetings or voice notes for quick capture. It would even be cool to take voice recordings from Telegram voice notes or WhatsApp.
13 Comments
Sign in to comment
Yes! Would love a Granola.ai style notetaker along with this!
+1
Great thing would also be, if the use could decide wheater to use local processing or cloud.
Yes, maybe the option to turn WhatsApp voice messages into text automatically? Including support for Hebrew Voice Messages?
Any chance?
TYVM IN ADVANCE
Can you please give some feedback on how this feature will be implemented? Will this just be a file-to-text transcription integration? Or will this be more in line with Notion's AI notes where you're able to record a meeting and it summarizes your meeting?
This is now actively worked on. A first version will go out to Capacities Believers soon.
I It would also be super convenient and efficient if what is dictated was entered In the correct blocks of the object through AI (i.e., like the function in TANA)
Agreed. I would also like to see the text to speech feature in capacities. And current speech to text dictation on ios appears to be quite unstable in caps. When I use it, text gets lost, although works in obsidian very successfully.
I would really love to see this feature brought to life in Capacities. Having tried both Mem and Reflect and both support dictation capture/transcribing and AI refinement... This really is a game changer. I'm able to speak with more freedom and complexity than I can ever type fast enough to capture. This feature gives you the ability to really let loose with thoughts and then form it into something more permanent and clean for your PKM.
I love the way Notion has implemented this in their 'meeting notes'. Essentially, it just records, input and output, regardless of what's going on. That way, if I'm having a WhatsApp call, or have to jump back and forth between different platforms, it just records. Same with mobile, regardless of what app is being used, it just records the input/output without any bells or whistles or fanfare.
I am excited to see this on the roadmap, but hope that the team can add the capability to the Linux AppImage as well as Windows and Mac clients. This would be unique for Linux and put Capacities way beyond any competition.
I personally am keen to break from being forced to use Big Tech and Capacities' model of local-first is very much in keeping with this. I hope the Linux client will continue to be an equal partner in the Capacities world.
a good workaround on the Mac is turning on dictation but it's not quite as flexible because it doesn't run in the background if you're switching tabs input fields, etc. having this would be a game changer. It also seems like the future of human computer interface to use a lot more voice and then let the AI rewrite summarize make bullets, etc.. It's a lot faster. I wrote this comment using Mac built-in dictation!
We released audio recording and analysis recently! It's currently available for all Believers and will be released to Pro soon. :)