Micro.blog

mako@sanjuans.life
mako@sanjuans.life

@jarrod hey — let me know if interested; I took your shortcut to make audio from text (heydingus.net/shortcuts/make-s), and changed it to make a call into OpenAI’s TTS service… output is awesome.

mako@sanjuans.life
mako@sanjuans.life

@jarrod icloud.com/shortcuts/5be340a30

just provide your own API key.

jarrod
jarrod

@mako Oh, very cool — thank you for sharing! I've been thinking about other improvements and might include this as an option if that's okay? (I'll have to figure out a workaround for the 4096-character limit.)

mako@sanjuans.life
mako@sanjuans.life

@jarrod of course!

here's the other thing that I hacked up -
- Find me 3 articles in my feed reader I didnt read yet.
- Get summaries for each via Kagi's Summarizer ( kagi.com/summarizer/index.html )
- Pass the summaries into ChatGPT, create me a 2 sentence overall summary, and then
- Pass all the results and create me an audio output to listen to later.

End result - a 3 min personal audio clip of articles I may find interesting.

mako@sanjuans.life
mako@sanjuans.life

@jarrod maybe call into ChatGPT (summarize to 4000 characters :))

also - @omnivore mobile app has their TTS already in there, streaming audio returned..

jarrod
jarrod

@mako Wow, heck of a job! That's very close to my idea of a RSS-catcher-upper. With a few more steps and integrations, you could totally turn those summaries into a self-hosted podcast to listen to in any podcast app. You've got my mind a-turnin'!

jarrod
jarrod

@mako Yeah, Omnivore's TTS is best in the business. I send long articles there that I want to listen to. That's a little different than what I'm using the shortcut for, though. My current experiment is to make an audio version of my blog posts to host alongside them on my site. So I'd want them to be exact, not summarized.

mako@sanjuans.life
mako@sanjuans.life

@jarrod yup; almost think can just script it up to run at 5am and have a summary ready for you when you wake up.

mako@sanjuans.life
mako@sanjuans.life

@jarrod yup! Omnivore just uses OpenAI’s service. Maybe you can bust past the 4096 limit by chunking the output…

In reply to
jarrod
jarrod

@mako That's what I'm thinking! Playing around with it now...

mako@sanjuans.life
mako@sanjuans.life

@jarrod and wow -- i love the idea of just dropping the audio into a "personalized" podcast feed.

jarrod
jarrod

@jarrod @mako I got stuck trying to figure out how to combine the multiple audio files when it's split up.

mako@sanjuans.life
mako@sanjuans.life

@jarrod are you trying to do this on iOS? May be easier just to use python to stream to a file.

jarrod
jarrod

@mako Yeah, ideally I'd be able to do it from anywhere. Maybe not possible at the moment.

mako@sanjuans.life
mako@sanjuans.life

@jarrod That’s my usual experience with shortcuts unfortunately. It gets me 90% there and then I have a battle to the death for the last 10%.

mako@sanjuans.life
mako@sanjuans.life

@jarrod I guess you could always try calling it from Pythonista / shortcut..

jarrod
jarrod

@mako I hear you. The editing experience has gone severely downhill too. Which is disappointing because it's my favorite "programming" environment.

jarrod
jarrod

@mako That's a thought. I've never tried Python or Pythonista, but always up for learning something new!