[TTS] Riva support Fastpitch + GST (global style token) model?

AI & Data Science Deep Learning (Training & Inference) Nemotron Speech + Riva
1

I want to train 1 model Nemo TTS Fastpitch + GST. I don’t know if Riva supports this model or not?

I found this model code on the fastpitch gst branch on the Nemo github that the team developed GitHub - NVIDIA/NeMo at fastpitch_gst. Can I training model using this code?

Thanks!
Hardware - GPU (A100/A30/T4/V100)
Hardware - CPU
Operating System
Riva Version
TLT Version (if relevant)
How to reproduce the issue ? (This is for errors. Please share the command and the detailed log here)

2

Hi @user63286

Thanks for your interest in Riva

I will check with my team and let you know

Currently for custom models, we support
FastPitch + HiFi-GAN IPA
FastPitch + HiFi-GAN ARPABET
FastPitch + HiFi-GAN LJSpeech

Please find the reference link below
https://docs.nvidia.com/deeplearning/riva/user-guide/docs/tts/tts-custom.html

Thanks

3

HI @user63286

I have inputs from the internal team

That branch is out-dated. The most up-to-date version of the work linked is https://github.com/NVIDIA/NeMo/pull/5499.

The Team is internally evaluating it, but we have no timeline or commitment on adding it to Riva at the moment.

Thanks