OpenTX sounds
TODO: merge with /arf-drone/docs/sounds
The OpenTX sound files are built by the Python code in opentx/.../radio/util - the main files being tts.py and tts_en.py.
If you look at tts.py you can see that it can use a number of speech synthesizers:
- SAPI and SAPI2 - the Microsoft Speech API.
- eSpeak - a very mechanical sounding open source synthesizer.
- Google - co-opts the speaking functionality of the Google Translate web service.
Mac voices
RCDiy provides an alternative on Github - OpenTXSounds - that can use the Mac voices. See SoundPackGenerator.sh for the script that actually generates the sound packs.
This Bash script works off the CSV files that are generated by the original OpenTX Python tty.py code - so you need the original CSV file for your language from the official OpenTX sound pack - the 2.2 sound packs are here.
The Mac voices apparently come from Nuance the creators of Dragon. You can download higher quality versions of the existing voices along with additional voices. Go to System Preferences / Dictation & Speech and select Customize... from the System Voice dropdown. I selected Ava, Allison, Samantha and Tom, in addition to the already selected voices, as I read that these are very high quality voices. Doing this triggers not just the download of these voices but also higher quality downloads of some of the already selected voices.
The voice data can be found under /System/Library/Speech/Voices and is substantial.
Ava, Allison and Samantha are probably my favorite voices. Allison despite being listed as an American voice doesn't sound very American.
$ say -vAva "stabilize, altitude hold, position hold, return to launch, land, otto"
$ say -vAllison "stabilize, altitude hold, position hold, return to launch, land, otto"
$ say -vSamantha "stabilize, altitude hold, position hold, return to launch, land, otto"
Why have I writen "auto" as "otto" above. The Ava and Samantha voices seem to handle "auto" fine but, for whatever reason, the Allison voice sounds very strange saying "auto" on its own:
$ say -vAllison "auto"
Preceding it with another word (rather than a comma) seems to improve things:
$ say -vAllison "and auto"
Or you can just try and help it out:
$ say -vAllison "awe toe"
$ say -vAllison "otto"
"Otto" seems to produce an identical sound to the American English "auto" sound.
Siri voice
Susan Bennett is the voice behind Samantha and Siri. Compare Samantha with the same sound from a sound pack created with Siri's voice and you'll see they're identical:
$ curl -O https://raw.githubusercontent.com/dale3h/taranis-siri-sound-pack/master/en/system/tada.wav
$ open -a "quicktime player" tada.wav
$ say -vSamantha "welcome to open t x"
And I confirmed, using the nearest to hand iOS device, that this is how Siri sounds on a real device (by default she reads a little faster than normal speed).
Note: many of the WAV files have been renamed in recent releases - tada.wav is one of them, it's now called hello.wav.