Provides GoogleTTS speeach synthesis for use with Snips AI.
Installation comprises 4 steps:
- Install dependencies
- Get SnipsGoogleTTS scripts
- Prepare the environment for the google API
- Tell Snips to use SnipsGoogleTTS.
The scripts depend on some installed software:
- git for installation
- curl for sending the POST requests to the REST API
- base64 for decoding the audio file retrieved from Google (in the coreutils package):
sudo apt-get install git-core curl coreutils
The scripts can be stored anywhere on the computer running Snips (most
probably a Raspberry Pie). A good location is for example
/opt/SnipsHelpers/SnipsGoogleTTS.
To install, just change to the target directory and clone the
repository:
cd /opt
sudo mkdir SnipsHelpers
sudo chown <YOUR.USER.NAME> SnipsHelpers
git clone [email protected]:andreasdominik/SnipsGoogleTTS.git
Verify that the two scripts say and saySnips are exectuable
(permissons -rwxr-xr-x) and all files and directories are readable
(three rs) and the directory Cache must be writable for everybody
(three ws).
ls -l should read somehow like:
drwxrwxrwx 4 andreas andreas 4096 Mär 13 10:03 Cache
drwxr-xr-x 2 andreas andreas 4096 Mär 13 10:03 Credentials
-rw-r--r-- 1 andreas andreas 7653 Mär 13 10:03 LICENSE
-rw-r--r-- 1 andreas andreas 2988 Mär 13 10:03 README.md
-rwxr-xr-x 1 andreas andreas 762 Mär 13 10:03 say
-rwxr-xr-x 1 andreas andreas 947 Mär 13 10:03 saySnips
drwxr-xr-x 2 andreas andreas 4096 Mär 13 10:03 src
If necessary change the permissions as:
chmod 755 say saySnips
cdmod -R a+r *
chmod -R 777 Cache
If you like, you can add a symbolic link to the say-executable from a directory in the
path, such as /usr/local/bin:
cd /usr/local/bin
sudo ln -s /opt/SnipsHelpers/SnipsGoogleTTS/say
You can test the intallation by calling say, as:
say en Hello I am snips. How are you?
The first argument is the language code; all other args will be concatenated as text to be spoken.
All generated audio files will be added to the cache. Therefore the say-command
can be used to explicitly add sentences to the cache (Snips will make
no on-line requests for already cached sentences).
If you want to change the voice for a language (this can be configured in the
script ttsREST.sh), the cache for this language must be deleted.
Go through Google's tutorial Quickstart: Using the command line.
In summary ...
- a Google Cloud Platform Project is needed,
- the Cloud Text-to-Speech API must be enabled and
- the JSON-file with the credentials must be downloaded.
Save the json file with the credentials in the Credentials sub-directory
of SnipsGoogleTT and rename it to google-credentials.json.
As last step the Cloud SDK must be installed. To check the installation run
$ gcloud auth application-default print-access-token
It should print an access token, which can be uses to access the Cloud Text-to-Speech API.
Edit the section [snips-tts] of the file /etc/snips.toml and
change the two lines defining the custom tts settings as shown:
[snips-tts]
## Choose one tts provider (defaults to picotts)
# provider = "picotts"
# provider = "makerstts"
provider = "customtts"
## customtts specific configuration (here configured to use picotts using the en-US language)
## available placeholder variables : %%OUTPUT_FILE%%, %%LANG%%, %%TEXT%%
customtts = { command = ["/opt/Snips/GoogleTTS4Snips/saySnips", "%%OUTPUT_FILE%%", "%%LANG%%", "%%TEXT%%"] }
Please be aware that the Google-Text-to-Speach runs on Google's servers and no longer locally on your Raspberry Pi. All text sent to the TTS is shared with Google (only once, because of caching).
Do not install this scripts, if you want to keep everything private!
Be aware that the Google Cloud services are not for free. However costs for the TTS are managable (mostly in the range of some Cents per year).
eof.