Tweetrad.io is a fun little side project that converts twitter feeds into audio using text to speech.
We use a distributed network of converter bots to handle the work queue. Jobs are placed on the queue via api calls from the javascript client that interacts with twitter's search api directly in order to avoid rate limiting of a centralized search server. Our app server is a lightweight Sinatra service that handles job queueing. We stream the audio directly from our nginx web server as they become available via the conversion process.
We are trying to decide if this app has a real use case or if it's just a fun thing. One of the ideas we are thinking about is an iphone app for keeping up with your tweet stream on the go. We would love to hear your feedback.
An iPhone app would be great. Especially if it could read tweets from your followers between songs, though I'm not sure if Apple gives you enough control over the iPod to do that.
That would be a fun idea. I'd probably buy it right away and start using it more then Tweetie (leave it open in the background while doing other work and just passively hear tweets).
Incase anyone does want to do this, Apple doesn't directly tell you when a song is starting/stopping, but they do tell you how long the song is (with MPMediaItem and the key MPMediaItemPropertyPlaybackDuration) and how far into the song you are (with MPMusicPlayerController).
To add why, it's rare that I'd want to listen to something when it's quicker for me to scan. The only time I'd do that is if it's inconvenient for me to read, ie when I'm on the move or driving. Then, I'd like for it to be fed to me audibly.
Good job getting it up and running. What do you use for text to speech conversion?
I like it too, zen, especially the female voice.
Would be neat to add special effects in the audio stream, e.g. if the tweet is about "happy hour" or "party", or if it's news you could have background clip like CNN.com headlines
"BreakingNews is excited to tell everyone that At least 17 people killed, 45 others injured after a passenger train collides with a freight train near Pakistan's Karachi."
I love it. We generate weather updates, which I'm used to hearing read by automated voices anyway. Now it's like we have our own staff of robots to do the same: http://tweetrad.io/?q=from%3Astormpulse
* tried to check out the about section of your site, but the feed kept moving the footer down. Maybe lock the footer in place or put that info to the side of the feed.
* A few repeated/stuttered sections. I heard this a few times during reads, but there didn't seem to be any specific thing that caused it.
* Speech cuts off after :
Feature requests:
* Log in with twitter account. Then I can hear my friend updates.
* Customization options> voicing, announcements (says, wants it to be known, is excited... Sometimes they can get annoying)
* Pause button
* Literal reads (instead of says to @xxx, it could say ... says 'ar tee at xxxx, blah blah blah'
All in all I really like this site. It's beautifully done and is actually quite entertaining.
Thanks for the great feedback. I'm Looking into the bugs.
As for the feature requests, logging in to view your personal feed and the pause button are on our roadmap and should be added soon.
You make a good point about the lead-ins. We initially had a bunch of silly lead-ins, but we've been getting consistent feedback that we should just keep it simple. We've already trimmed down the lead-ins significantly. Sounds like we should keep trimming.
Very interesting project and I had fun listening through the recent tweets from HackerNews and ShitMyDadSays. A minor annoyance is the robot voice, but I realize you may be at the mercy of the text to speech engine on that one. My only criticism would be when reading @replies. It says the name of the person you're replying to twice in quick succession and it's hard to understand. I'd either say it once or just put a pause in there. ;)
This application has the potential to provide an enjoyable, useful ambient experience. Personally, I can imagine myself having this on in the background, while I'm working. Before I would consider doing that, though, the speech quality needs to be improved.
Thank you all for the great feedback. We will be replying to some of your questions. Also the How It Works, blog post is in progress. We're currently featured on the mashable.com site so we are making sure the bots continue crunching those tweets.
I assume the "music" in between readings is supposed to sound like the radio is tuning for a new station? If this is the case, it's a neat idea, but it was a little frustrating because you want to hear the music, but it's not clear - although I'm not sure on any better suggestion (i.e. elevator music, muzak..lol).
brk 23 has a good point about filters, that would definitely make it more usable.
Nice work! Haven't seen any other fun web integration of text-speech like this (I definitely don't count dictionary.com pronouncing words for me as fun!)
I like it. Love the transition audio, reminds me of Bioshock. Listening as I type this comment and it's pretty cool. Not sure how often I would use it or if I would use it regularly at all. Does it push live updates?
Agree with the suggestions mentioned already (mute, remove extra commentary).
I'm impressed with the coordination and technology that went into this.
Kinda says something about America though when not only are we too lazy to write full blogs, read full blogs, but now we don't even want to read our tweets. This of course is not a serious comment on your project as more as it is commentary on America's priorities.
Great idea! It looks great too! I second (third, fourth, whatever) the calls for a mute button. I'd also like to see the ability to change the voice being used. I noticed that between channels it's different, any way to choose between those voices?
its a very neat project. As said before I'd also love to see how you put it together. Also, maybe you can put a little toggle button next to tweets that people can click when they think a word is not said correctly.
I don't know if it has any real business potential, but it's certainly an interesting little project. I'd love to see a blog post about how you put it together.
We use a distributed network of converter bots to handle the work queue. Jobs are placed on the queue via api calls from the javascript client that interacts with twitter's search api directly in order to avoid rate limiting of a centralized search server. Our app server is a lightweight Sinatra service that handles job queueing. We stream the audio directly from our nginx web server as they become available via the conversion process.
We are trying to decide if this app has a real use case or if it's just a fun thing. One of the ideas we are thinking about is an iphone app for keeping up with your tweet stream on the go. We would love to hear your feedback.