CoolFace
Modelpublic

alakxender/mms-tts-div-finetuned-md-f03

sourceHugging Facemitupdated 1y agoView on Hugging Face
0likes39downloads
Model Card

Divehi TTS – Female Voice (VITS-based)

This is a fine-tuned VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) model for Divehi speech synthesis. The model produces Female voice audio from Thaana-scripted Divehi text. Fine-tuned from Meta’s MMS-TTS architecture using a curated dataset of synthetic Divehi speech.

Model Details

FieldValue
Model IDalakxender/mms-tts-div-finetuned-md-f03
Base ArchitectureMMS-TTS (VITS)
LanguageDivehi (dv)
VoiceFemale
Sampling Rate16 kHz
TokenizerVITSTokenizer
Inference EngineTransformers (🤗 Hugging Face)

Usage

python
from transformers import VitsModel, VitsTokenizer
import torchaudio

tokenizer = VitsTokenizer.from_pretrained("alakxender/mms-tts-div-finetuned-md-f03")
model = VitsModel.from_pretrained("alakxender/mms-tts-div-finetuned-md-f03")

text = "މޫސުން ވަރަށް ގޯސްވެ، ފުވައްމުލަކުން ފެށިގެން އައްޑުއަށް އޮރެންޖް އެލާޓް ނެރެފި"
inputs = tokenizer(text, return_tensors="pt")
waveform = model.generate(**inputs).waveform[0]

torchaudio.save("output.wav", waveform.unsqueeze(0), 16000)

Evaluation Summary

  • —Model: alakxender/mms-tts-div-finetuned-md-f03
  • —Evaluated Samples: 3
  • —Avg Estimated MOS (UTMOS): 2.102
json
  {
    "5": "Excellent (very natural)",
    "4": "Good (mostly natural)",
    "3": "Fair (some robotic quality)",
    "2": "Poor (noticeably unnatural)",
    "1": "Bad (unintelligible or very synthetic)"
  }
  • —Artifacts:
  • —🎵 Audio: outputs/audio/
  • —📊 Spectrograms: outputs/spectrograms/
  • —📄 Report: outputs/report.txt
  • —📈 MOS Scores: outputs/mos_scores.txt

Acknowledgements