You can deploy this model and start earning money today!

Axiveri/
Renpiper-pcm-V1

Name: Renpiper-pcm-V1
Rating: 2.5 (19 reviews)

Views

No views yet

Renpiper-pcm V1

This is a repackaged, safetensors-normalized mirror of Meta AI's facebook/mms-tts-pcm checkpoint (Nigerian Pidgin / pcm, VITS architecture), part of Meta's Massively Multilingual Speech (MMS) project. The weights are unmodified -- no fine-tuning has been applied. This repo exists purely to give a clean, single-folder, safetensors-only layout for downstream loading.

What this is NOT

Not an original or independently trained model.
Not fine-tuned or adapted for any particular use case beyond what facebook/mms-tts-pcm already does.
Not available for commercial use or commercial relicensing under any name -- see License below.

License

CC-BY-NC 4.0, inherited unchanged from the source checkpoint. Non-commercial use only. This means this repo cannot be used as the basis for a paid product, client deliverable, or commercial deployment (including under a different product name) without a separate commercial license from Meta AI.

Citation

If you use this model, cite the original MMS paper:

@article{pratap2023mms,
  title={Scaling Speech Technology to 1,000+ Languages},
  author={Vineel Pratap and Andros Tjandra and Bowen Shi and Paden Tomasello
    and Arun Babu and Sayani Kundu and Ali Elkahky and Zhaoheng Ni and
    Apoorv Vyas and Maryam Fazel-Zarandi and Alexei Baevski and Yossi Adi
    and Xiaohui Zhang and Wei-Ning Hsu and Alexis Conneau and Michael Auli},
  journal={arXiv},
  year={2023}
}

Model developed by Vineel Pratap et al., Meta AI. All credit for the underlying model belongs to Meta AI / the MMS project, not to this repo's maintainer.

Inference

python

import torch
from transformers import VitsModel, AutoTokenizer

model = VitsModel.from_pretrained("Axiveri/Renpiper-pcm-V1")
tokenizer = AutoTokenizer.from_pretrained("Axiveri/Renpiper-pcm-V1")

text = "Your Nigerian Pidgin text here"
inputs = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    output = model(**inputs).waveform

import soundfile as sf
sf.write("out.wav", output.squeeze().cpu().numpy(), model.config.sampling_rate)