Uses voice generation by Uberduck.AI.
Works at least with L10s Ultra
, L10 Pro
, Z10 Pro
and W10
.
Optimized for valetudo use.
Current voice packs:
In Valetudo go to "Robot Settings" -> "Misc Settings"
Enter the following information in the "Voice packs" section:
CUSTOM
- or anything but language codes like EN
GLADOS
HASH.txt
file
bc10b417ace1dd1aaf4f3e1525aa3b73
Click "Set Voice Pack"
Check if the *.ogg
files have been copied to /data/personalized_voice/CUSTOM/
directory on the robot.
If not, extract the voice.tar.gz
and copy the *.ogg
files using your favorite scp
tool.
ffmpeg
and vorbis-tools
sudo apt install ffmpeg vorbis-tools
pip install uberduck
keys.csv
kirby-tiff
with one of the voices)python generate.py --voice kirby-tiff [--volume 2] [--normalize]
The default voice is glados
.
Note: Normalization uses ffmpeg
and may create random artifacts in the resulting audio. It is not recommended with Uberduck.AI voice characters.
Note: voice list can be found at https://app.uberduck.ai/quack-help . Not all voices are good. Please, consult: https://app.uberduck.ai/leaderboard/voice
Listen to the resulting .ogg
files for any artifacts. Change the text if necessary.
The defaults are loaded first and are found in the defaults
folder.
Then the values in each character folder override the defaults.
Each folder can have following files:
./sayings.csv
- a list of saying IDs and saying texts. If the text is empty, generator will look for a prepared .ogg
file../replacement.csv
- a list of words and their replacements, in case the AI characters cannot pronounce those correctly../*.ogg
- prepared ogg audio files, used when the text for the saying ID is empty.In each voice-character folder:
./voice.tar.gz
- archive of the voices generated../HASH.txt
- contains the md5 hash code of the voice archive../ogg/*.ogg
- resulting ogg audio files, that go into the archive../tts/*.wav
- transformed text audios with name being the hash of the text.Please, don't add any obscenity or racist stuff. We cannot accept this on a public directory. There is no point in offending anybody. Reserve this stuff for your local installation.
Also, please, don't add anything copy-protected.
Thanks to https://github.com/ccoors/dreame_voice_packs and https://github.com/Findus23/voice_pack_dreame for the inspiration and https://uberduck.ai for the voice generation.