You can deploy this model and start earning money today!

SharadhNaiduTrains/
sravaani-flow-model

Name: sravaani-flow-model
Availability: InStock

transformers

sravaani_tdt

feature-extraction

automatic-speech-recognition

custom_code

mirror

Views

No views yet

SraVaani-1.0 — unmodified mirror

This is a byte-for-byte mirror of ARTPARK-IISc/SraVaani-1.0.

All credit belongs to the original authors at ARTPARK and the Indian Institute of Science. No weights, code or configuration have been modified.

Why this mirror exists

It is used by SraVaani Flow, an offline dictation app, so that people evaluating the app on a shared machine can install it without each needing their own Hugging Face account and access token.

The original repository is gated (auto-approved on accepting the terms). This mirror is redistributed under the MIT licence that the original release grants. If you are using the model for your own work, please use the official repository — it is the canonical source, it is where updates land, and the authors ask that users share contact details so they can gauge adoption:

https://huggingface.co/ARTPARK-IISc/SraVaani-1.0

If the ARTPARK-IISc authors would prefer this mirror not exist, contact vaanicontact@gmail.com or open a discussion here and it will be taken down.

About the model

A ~430M parameter FastConformer ASR model with a hybrid TDT-CTC decoder, supporting 65 Indian languages and dialects. It ships as an exported TorchScript graph, so inference needs only torch, transformers and sentencepiece — no NeMo toolkit.

Files

File	Size	Purpose
`model-asr.fp16.ts`	867 MB	TorchScript encoder + decoder/joint graph, fp16
`tokenizer.model`	318 KB	SentencePiece vocabulary (5000 tokens)
`preproc.pt`	132 KB	Mel filterbank, STFT window, feature parameters
`config.json`, `model_meta.json`	—	Model configuration
`modeling_sravaani.py`	—	HF wrapper (needs `trust_remote_code=True`)
`configuration_sravaani.py`, `processing_sravaani.py`	—	Config and processor
`example_inference.py`	—	Minimal usage example

Usage

python

import torch
from transformers import AutoModel

model = AutoModel.from_pretrained(
    "SharadhNaiduTrains/sravaani-flow-model",
    trust_remote_code=True,
    dtype=torch.float16,
).to("cuda").eval()

# note the list: a bare ndarray would be iterated element-wise
print(model.transcribe(["sample.wav"])[0])

Licence

MIT, inherited unchanged from the original release.

MIT License

Copyright (c) 2026 ARTPARK, Indian Institute of Science

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Citation

Please cite the original authors:

bibtex

@misc{pulikodan2026sravaani10scalinginclusive,
      title={SraVaani 1.0: Scaling Inclusive Speech Recognition for Indic Languages},
      author={Sujith Pulikodan and Agneedh Basu and Pavan Kumar J and Pranav D Bhat and Suryansh Shukla and Nihar Desai and Prasanta Kumar Ghosh},
      year={2026},
      eprint={2608.08235},
      archivePrefix={arXiv},
      primaryClass={eess.AS},
      url={https://arxiv.org/abs/2608.08235},
}