nemo-balochi-g2p-conformer

Balochi Conformer G2P

Balochi Conformer G2P converts Pakistani Balochi text written in the Perso-Arabic script into broad IPA-style phoneme strings. It is a character-level NVIDIA NeMo Conformer CTC model intended for TTS preprocessing, pronunciation-label bootstrapping, and speech-dataset quality checks.

First Public Balochi G2P Release

To the best of our knowledge, this is the first publicly released model dedicated to Balochi grapheme-to-phoneme conversion. At the time of release, no open Balochi-specific G2P model or comparable public model resource could be found. This release is intended to establish a practical starting point for a language that remains severely under-resourced in speech and language technology.

This model belongs to the Pakistani Languages G2P Family alongside the Pashto Conformer G2P and Sindhi Conformer G2P.

Architecture

• Framework: NVIDIA NeMo
• Objective: character-level CTC grapheme-to-phoneme conversion
• Encoder: 8-layer Conformer
• Hidden size: 192
• Attention heads: 4
• Convolution kernel: 15
• Output: broad IPA-style phoneme sequence
• Default source expansion: each non-space input character is repeated three times for CTC alignment
• Maximum expanded source length: 1024 characters

Quick Start

Install the dependencies:

pip install -r requirements.txt

Run from a cloned repository:

python inference.py --text "گلے چاروں کا مئے دعا ہمیشے اللہ پاک تمام مسلماناں عمرہ حج کنگے توفیق بدات آمین آمین"

Example output:

guleː t͡ʃaːroːkaː maeː d̪uaː hameːʃeː allaː paːk t̪amaːm musulmaːnaː umraː had͡ʒ kanageː t̪oːfiːq bid̪aːt̪ aːmiːn aːmiːn

Another example:

python inference.py --text "ہاں چتورے؟ شکر آ۔ تو گانوکانی وڑا جنگلا سرگردانے؟ تو عالیہ سئی ہے؟ عالیہ؟ چے عالیہ؟"

hãː t͡ʃat̪oːreː ʃukr aː t̪oː gaːnoːkaːniː waɽaː d͡ʒaŋgalaː sargardaːneː t̪oː aːliːjaː saiː eː aːlijaː t͡ʃeː aːlijaː

Or let the script download the model from Hugging Face:

python inference.py \
  --repo-id TBOGamer22/nemo-balochi-g2p-conformer \
  --model-path missing.nemo \
  --text "گلے چاروں کا مئے دعا ہمیشے اللہ پاک تمام مسلماناں عمرہ حج کنگے توفیق بدات آمین آمین"

For multiple inputs, save one sentence per line in a UTF-8 file:

python inference.py --text-file inputs.txt --json

Model Files

The public repository intentionally contains only inference files. It does not include datasets, data splits, manifests, prediction tables, training logs, or trainer checkpoints.

Intended Use

Use this model to obtain broad IPA-style pronunciation sequences for Pakistani Balochi text, especially as a preprocessing component for speech synthesis and corpus tooling. Review predictions before using them as authoritative linguistic annotations.

Limitations

• Balochi has dialectal and orthographic variation; a single spelling can have more than one valid pronunciation.
• The output is broad, model-predicted IPA rather than narrow phonetic transcription.
• Code-switching, uncommon spellings, foreign names, punctuation-heavy text, and unseen characters may reduce reliability.
• The model is not a substitute for review by Balochi speakers or trained linguists.

License and Author

Released under the MIT License.

Author: Talha Bin Omar

Citation

@misc{omar2026balochig2p,
  author       = {Talha Bin Omar},
  title        = {Balochi Conformer G2P},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/TBOGamer22/nemo-balochi-g2p-conformer}}
}