The Future of Speech Text APIs

4min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

The reason why whisper.cpp doesn't run very quickly on iPhones is because Apple has not done their work to optimize that one the way they did with the stable diffusion. If Apple actually wanted to offer a really great dictation to speech API or to speech text API, they could massively improve the API by using this new kind of technique. That's the kind of thing that I think is more likely for Apple to offer with these new AI techniques sooner than having them offer some kind of full blown customer facing quote AI powered feature.

Pre-show:
- 🗣️ New Member’s Special posted 🗣️
- A New Yorker’s View of the United States
Follow-up:
- The Kobo FU continues
  - The Libra 2 does have support, kinda (via Stefano Costatini)
  - Install NickelMenu
  - Add the Dropbox menu item
    - Libra 2
    - Other Kobos
- Multiple Displays on M1/M2 Macs
  - Matthaus Woolard points out there are only 2 display controllers
  - Hector Marcan has now-deleted tweets covering this (via Richard Stephens)
  - Edward Munn (among others) clues us into a DisplayLink-based workaround
  - Wavlink USB 3.0 ↔ HDMI Adapter (via Steve McWee)
  - Lior Shakedd summarizes the drawbacks
- The woes of the iBook G4 (via Ryan Maxwell)
  - Screen Spanning Doctor
  - Compatibility List
- Final Cut Pro on iPad
  - Some M1 Pros had 16GB RAM and virtual memory (via Chris)
  - Final Cut Pro handles project files in ∞ ways (via Brendan)
- Dell UltraSharp 32” 6K Monitor is $3,200
  - Resolutions
    - XDR = 6016 × 3384
    - Dell = 6144 × 3456
    - Dell has an 8K version available
    - Dell offers a variety of sizes
  - HDR600 only
WWDC Vibe Check 🧘🏻‍♂️
- Avoiding Copland 2010
- You Ruined Everything
Big watchOS upgrades coming?
Post-show: How should Casey price Callsheet?