Replace mimic3 with piper #686
Conversation
There was a problem hiding this comment.
A few things:
- How is the model downloaded?
- I think we should implement caching so the tts is not calculated every time (we had this in mimic3)
- With mimic we had this seperate server that was started by systemd (configured in Ansible) that way we could e.g. announce the IP etc. at startup
| @@ -229,3 +229,6 @@ doku/* | |||
| **/workspace_status.json | |||
|
|
|||
| .pytest_cache/ | |||
|
|
|||
| # tts model | |||
| */bitbots_tts/model/* | |||
There was a problem hiding this comment.
I would suggest **/bitbots_tts/model/ to ignore the whole directory at any depth (in case we move it at some point).
| @@ -16,6 +17,12 @@ | |||
|
|
|||
| from bitbots_msgs.msg import Audio | |||
|
|
|||
| # Load the Piper voice | |||
| bb_tts_dir = Path(__file__).parent.parent / "model" # TODO: check how to get nice relative paths | |||
There was a problem hiding this comment.
I mean depending on where you wanna locate the model folder you should probably use:
get_package_prefix, get_package_share_path which would give you e.g. ~/colcon_ws/install/share/bitbots_tts.
The current solutions should also work, but I would then use an absolute path to ensure that it will work when building without symlinks.
| } | ||
| with io.BytesIO() as buffer: | ||
| with wave.open(buffer, "wb") as wav_file: | ||
| voice.synthesize(text, wav_file, **synthesize_args) |
There was a problem hiding this comment.
This would mean that we will generate a new wav file each time?
mimic was implicitly caching the already generated voices, which was also the reason for the web server if I remember correctly.
Have you tried how long the wav generation takes?
There was a problem hiding this comment.
It is real time on my laptop, but I did not try it on the robot yet. Maybe we should cache them again though... just to be sure
Summary
fixes #530
Proposed changes
Related issues
Checklist
colcon build