search

dusty-nv / jetson-containers

Machine Learning Containers for NVIDIA Jetson and JetPack-L4T
MIT License
1.9k stars 416 forks source link

Add Dockerfile for piper-tts #468

Closed michaelgruner closed 3 months ago

michaelgruner commented 3 months ago

@dusty-nv added a test. Looking good:

Inference duration: 2.096s
Audio duration: 39.114s
Real time factor: 0.054
Piper TTS: OK

I'm unsure how to proceed with the readme and all the config and build stuff. I fyou give me some pointers I can get that done too.

dusty-nv commented 3 months ago

Thanks @michaelgruner, no worries, I will take care of the other stuff and getting the images on dockerhub 👍

dusty-nv commented 3 months ago

FYI in f1c6eae I just moved piper-tts under jetson-containers/packages/audio with the other ASR/TTS containers (it is still in the dev branch until I get the images pushed)

Thanks for the nice test.py! Will piper automatically download the referenced onnx (en_US-lessac-high.onnx) from HuggingFace, or does one need to manually download them? If the later, I will add the auto-downloader from the huggingface_hub container (the huggingface-downloader subshell tool from that is what I use repo-wide for keeping models on huggingface cached/organized, and will output your local path to the cached model after it downloads it)

michaelgruner commented 3 months ago

It is auto-downloaded if not found.

BTW, one detail is that I left the piper and piper_phonemize at /. Not sure if you want to remove them