Notices

So, I've been looking into other speech synthesis platforms, because eSpeak is very robotic-sounding, especially compared to commercial efforts (google now, siri, cortana etc). As far as I can tell, the best alternative is flite. Flite's default voice is very robotic-sounding; however some contributed voices sound much more natural than eSpeak. I'm adding the best-sounding one I can find (cmu_us_clb) and packaging it as a separate package; if Saera sees it installed, it will give you the option to use it instead of eSpeak. (I'm not bundling it by default, because a, it's over 25 MB and b, I still need to bundle espeak anyway to generate phonemes for voice recognition of song titles, contact names etc.)

Here is a video I recorded, comparing eSpeak and flite side-by-side on the same outputs. Unfortunately I don't have a proper microphone, so it's just using the camera mic.

More work has been done on navigation. I've changed the algorithm to look for your position along a path, instead of merely triggering directions when you get to intersections. This allows for better accuracy, especially with closely-spaced turns, without sacrificing reliability. It also allows the next instruction to be triggered as soon as you turn, rather than waiting until you leave a certain radius of the turn.

Here is a video, showcasing several things:

• Internet voice recognition mode
• Local search
• Turn-by-turn directions
• flite speech synthesis

For navigation, some graphic info would be welcome too. A simple arrow and distance would help ("↱ 300m").

For TTS: have you seen (and heard) WaveNet?

These are already present; they show up on the right side of the screen, next to the direction text. I will consider making them more visible, though.

I have. Unfortunately, WaveNet is not available for download. There are many speech synthesis engines out there that sound better than espeak or flite, but they are all either paid engines, web APIs (which do not work offline and have latency in playback), or research projects like WaveNet which are not available for anyone to use.
Edit: Also, WaveNet is quite slow; on a laptop, it takes about 90 minutes to synthesize one second of speech. On a phone, it would probably be about 5 times slower still.

Trying out Saera again, after long absence. Not a clue what python-beautifulsoup does... How do you set alarms? Because "usual" alarms, set by the clock, do wake up the device even from complete shutdown. Also, cities to know time in, could be taken from the Clock's list, too, if it helps.

Thank you. Good day, good night, and good luck.

To set an alarm, say "Set alarm for quarter past nine" or "Wake me at seven twenty five" or something similar. It sets a native clock alarm, that should wake the device even from poweroff state. (On supported devices, that is; RTC alarms don't work on most non-Jolla SFOS ports.)

I'm gearing up for a big release tomorrow, and am hitting a snag on my Harmattan version: I recompiled Julius for harmattan, but after I reboot the phone it refuses to run the julius binary, saying "Permission denied" - even as "root". Permissions are still 755. Copying it to a file with a different name and then copying it back fixes it until next reboot. Anyone with Harmattan experience who might know how to fix this?

I'm releasing Saera version 2.3 for SFOS only now, to hopefully be followed by the Maemo and Harmattan releases in the next few weeks. What's new here:

• The updated navigation engine has finally arrived! This includes fixes such as looking at position relative to the path instead of going waypoint-to-waypoint for improved accuracy and less chance of losing the route.
• When Saera is speaking, any music playing is paused so that information can be clearly heard.
• Added a Settings page. This allows you to enable online voice recognition, which is disabled by default for privacy, but which when enabled improves recognition accuracy and vocabulary.
• Fixed display issues on high pixel density screens (Jolla C, OnePlus X, etc)
• Added support for flite (which is a separate package, because it's big)
• Fixed a lot of other miscellaneous bugs

A few things that I started working on did not make it into this release, as they are not finished yet and I wanted to get this released before the coding competition was over. These include more settings on the settings page, reading out text messages and possibly dictation. I expect to release those sometime soon though.

Anyone with the eternal initializing speech recognition problem got saera working (without a factory reset)?

Could you start saera from a terminal (run harbour-saera) and let me know what error you are getting?

I'm running into a dependency problem: espeak isn't found.

SFOS 2.0.5.6 Onyx (oneplus x) port. I've pkcon refresh(ed). I seem to recsll espeak was broken before in the 'crashes you device' sense--has it bren pulled from the repos? Is it too early in the morning for me and I'm forgetting something stupid?

Thanks