Based on some other projects I've been working on (BlueBell), I used the idea of rss-driven text to speech combined with cron job built on a Chumby Hacker Board, crammed it inside a cheap alarm clock and Viola! Enter the AlarmTock.
After the break you can here a quick sample of the output. It's not quite Tony Stark's 'Jarvis' yet, but it's a step in the right direction. Currently only does weather and top headlines, but I'm working on Facebook updates, personalized TV lineups, and todo lists.
(more after the break)
(embedded wav removed)
(here's the link to the wav, since the embedded is being flaky)
Sample Alarm
Here is the code written in python (not pretty yet...still playing):
For this to work you'll need to build the following from source:
mplayer
python
sox
Here's the build. All I did was wire the headphone hack of the Chumby directly to the alarm clock's speaker.
blue wire is just just headphone cord soldered to the alarm clock speaker |
dremeled out a hole for the chumby power supply |
...and a hole for a usb wifi adapter |
...tada |
Do some text replacements for Ex:
ReplyDelete50s > Fifties
MPH > Miles Per Hour
Etc, Etc, Etc...
Kinda wish i had a chumby to do this all with and pair it with some better voices, maybe even set it up like a news program with separate voices for the news caster and weather person, sports, traffic(well if there was a relevant RSS for my area which there isn't), and all that :)
Yep, text replacement is already on my list of todos.
ReplyDeleteThe TTS was the fun part. Open source TTS voice quality is kind of terrible, so what I did was hijack Google Translate TTS function (when you use google translate, there's an option to export the resulting translation to a wav file) but instead of translating it simply does english to english. I think the voice quality is far superior to the MBOLA voices I was trying and I'm not weighing down the Chumby with a heavy TTS program.