Access Whissle STT Hinglish-Loans on Hugging Face
This model is licensed for inference only — no training, fine-tuning, distillation, or reverse engineering permitted. Accept the license to access. Automatic approval.
By clicking "Agree", you accept the Whissle Inference-Only License Agreement. See the LICENSE file for full terms. Key restrictions: INFERENCE ONLY — no training, fine-tuning, distillation, model compression, or reverse engineering permitted. Free for inference use under 100M MAU. "Powered by Whissle" attribution required for redistribution.
Log in or Sign Up to review the conditions and access this model content.
Whissle STT Hinglish-Loans
Hindi-English code-mixed (Hinglish) speech recognition model optimized for conversational audio in financial and customer service domains. Built on Conformer-CTC architecture with a dual-head tag classifier that extracts speaker metadata and conversation intent in real-time.
Model Details
| Architecture | Conformer-CTC (EncDecCTCModelBPE) + dual-head tag classifier |
| Encoder | 512-dim, Conformer layers |
| Download size | ~478 MB |
| Format | ONNX (CPU and GPU compatible) |
| Sample rate | 16 kHz mono |
| Languages | Hindi, English, Hindi-English code-mixed |
Tag Classifier Outputs
The dual-head classifier runs on pooled encoder features and outputs five categories per utterance:
| Category | Classes | Labels |
|---|---|---|
| Age | 3 | CHILD_TEEN, ADULT, SENIOR |
| Emotion | 5 | NEUTRAL, HAPPY, SAD, ANGRY, FEAR |
| Gender | 2 | MALE, FEMALE |
| Intent | 13 | GREETING, IDENTITY_VERIFY, PAYMENT_REMINDER, PAYMENT_INSTRUCTION, CLAIMS_PAID, PROMISE_TO_PAY, PAYMENT_QUERY, AMOUNT_DISPUTE, FINANCIAL_HARDSHIP, COMPLAINT, URGENCY_PRESSURE, ACKNOWLEDGMENT, OTHER |
| Role | 3 | AGENT, CUSTOMER, OTHER |
Quick Start
Use with the Whissle STT Inference Server:
git clone https://github.com/WhissleAI/whissle_stt_inference.git
cd whissle_stt_inference
./setup.sh --model hinglish-loans
Or load directly with ONNX Runtime:
import onnxruntime as ort
session = ort.InferenceSession("model.onnx", providers=["CPUExecutionProvider"])
outputs = session.run(None, {"audio_signal": mel_features, "length": lengths})
License
Whissle Inference-Only License — inference only, no training/fine-tuning/distillation/reverse engineering. Free under 100M MAU.
- Downloads last month
- 108
Model tree for WhissleAI/STT-hinglish-loans-ONNX
Base model
nvidia/parakeet-ctc-0.6b