View profile

Hearing Voices - Issue #38, or Misunderstanding 5% of things you hear...is actually good

Revue
 
"[With audio production], you’re on a constant journey of keeping up with technology."- Andrew Mason,
 

Hearing Voices

December 17 · Issue #38 · View online
This week's news about voice computing apps, delivered directly to your inbox.

“[With audio production], you’re on a constant journey of keeping up with technology.”
- Andrew Mason, Descript (more on that below)
“[With voice technology], you can keep up by subscribing to Hearing Voices.”
- You? To your friends? About this newsletter?

On California time, so getting this edition out later than usual. Lots happening:
To Try:
> Descript – The new “audio word processor” from Andrew Mason, creator of Detour (cool audio tour app) and Groupon (some kind of couponing app, I think). He also talks on a podcast here about the future of the audio industry, it’s a fun interview.
> Storyline – Create voice apps without coding. This is especially cool because it automatically syncs with your alexa account and creates the models you need. 
> #Podcasters – A new directory of the best resources for podcast creators.
2 Listen 2:
> Me, on the voicebot podcast, talking funding and building voice tech startups. 
> Interesting to see Erik Torenberg’s new podcast Venture Stories do an exclusive window with the breaker podcasting app. Breaker also promoted their podcast prominently within the app.
> If you haven’t downloaded the breaker podcasting app, check it out. It’s getting really good. You can see what podcasts your friends are listening to (you can add me at breaker.audio/u/@matthartman)
To Read:
> Wirecutter updated its Amazon Echo vs. Google Home comparison. They pick the 2nd generation echo as the best product. Not sure I agree with this conclusion.
Mozilla’s open source voice recognition tool nears human-like accuracy. Their tool, called Deep Speech, has a 6.5% per word error rate (which is apparently very good). Thanks @pmontee for alerting me to this one.
Comparing the accuracy of automatic transcription services – written as part of the Descript launch, the TLDR is that speech detection is getting really good. Worthwhile read. 
But what if you misunderstood 5% of things you heard?!
For Devz:
I came across two data resources I thought were cool and that I hadn’t seen:
> A transcription of all ted talks
Carnegie Mellon Speech Consortium - Royalty-free source code for all components of a speech recognition system, including decoding, spoken language parsing, training of acoustic and language models, as well as robust signal processing.
> Mozilla common voice project - a web, android and iOS app for collecting speech donations for the Common Voice project.
Hiring
> Are you a developer interested in voice first? Want a job? Let me know and I’ll intro you to some companies looking to hire.
Have a great weekend everyone,
- Matt hrt.mn
Did someone forward this to you? Subscribe at hearingvoices.xyz
Follow me on twitter @matthartman
Support Hearing Voices by joining here and get timely alerts on voice news, data on audio-focused products, and chat with me about topics in voice UI.
Did you enjoy this issue?
If you don't want these updates anymore, please unsubscribe here
If you were forwarded this newsletter and you like it, you can subscribe here
Powered by Revue