Hi All,
Has anyone looked at Nvidia’s Jarvis framework? Not sure how open it is, but it may be good option for some of the back-end TTS and natural language processing - standalone server?
Blog post - Creating Voice-based Virtual Assistants Using NVIDIA Jarvis and Rasa | NVIDIA Developer Blog
It was presented as part of the Nvidia GTC 21 conference keynote. [S31895]
Yes, the demo shown in the keynote was quite impressive. I tried to experiment with some of the building blocks (e.g. Nemo TTS) but wasn’t successful with installation/build process. While some parts use well known technology, e.g. Tacotron for TTS, Nvidia bundles them into their Nemo/Jarvis eco system in way that locks you into that.
When interested in offline/standalone STT and TTS have a look at coqui.ai (former Mozilla TTS and DeepSpeech).