I've been playing with the new text-to-speech features, and they are SO much better than before (at least on Android). The new podcast-like audio interface is smooth enough that you could use this safely while driving, walking, or otherwise commuting. I am a heavy audiobook listener, and this is finally a viable option for catching up on articles that I want to read passively.

I thought the old voices were fine, both on Android and iOS. The new ones are from the Amazon Polly API (according to https://www.theverge.com/2018/10/11/17961564/pocket-redesign...). Between the two schemes, I think you trade off performance/battery (old TTS system) life with data usage (new version).

The new voices are maybe a little more natural sounding, but the pauses between sentences were better with the old Android voices, IMO. It looks like it will still use the system TTS engine when there's no data connection. This is cool, and I hope they make further improvements in the TTS area!

Seeking forward and backward left a lot to be desired on the old version. It skipped by paragraph, which worked but was not great since different articles have different paragraph lengths. In the app, you can now skip back and forward 15 seconds, which is more natural.

Unfortunately, on my Bluetooth headphones, skipping backward goes to the previous article, not back 15 seconds. They really need to have this as an option, because even with ANC headphones (Sony WH-1000XM2) sometimes you miss things due to loud noises or inattention, and you just want to listen to the last sentence or so. Audible and Google Play Books have this little feature that makes a huge difference.

Funny enough, I had a side project that was supposed to be a TTS-focused reading app, because I wasn't satisfied with the options from Pocket and Instapaper at the time. I built out the backend using a couple AWS Lambda Functions written in Node.js:

- One to strip an article to unformatted text only using the Mercury Web Parser API (https://mercury.postlight.com/web-parser/) and to do some additional cleanup. Text was saved in DynamoDB

- One to synthesize the text with Amazon Polly and save to S3. This was to be done when the user wanted to listen, not when the user saved the text article, to reduce costs

- One for streaming the article. This was where I left off. The idea would have been to get it to stream and sync up across your life, such as being able to pause on your phone then pick up on your Google Home or Amazon Echo. I never got around to writing any of the apps for the individual platforms, as I was focused on another project.

It's great to see Pocket create an almost-perfect listening app, so now I can focus on other stuff I want to do.