---
license: apache-2.0
language:
- en
base_model: openai/whisper-medium
tags:
- whisper
- automatic-speech-recognition
- legal
- supreme-court
- india
datasets:
- kirandevraj/supreme-court-hearings-asr
metrics:
- wer
- cer
pipeline_tag: automatic-speech-recognition
library_name: transformers
model-index:
- name: whisper-medium-supreme-court-hybrid
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: Supreme Court Hearings ASR (hybrid)
      type: kirandevraj/supreme-court-hearings-asr
    metrics:
    - type: wer
      value: 9.2
      name: Test WER
    - type: cer
      value: 5.2
      name: Test CER
---

# whisper-medium · Supreme Court hearings (hybrid dataset)

A **full fine-tune** of [`openai/whisper-medium`](https://huggingface.co/openai/whisper-medium)
(all 763M parameters) on the [`kirandevraj/supreme-court-hearings-asr`](https://huggingface.co/datasets/kirandevraj/supreme-court-hearings-asr)
dataset — 46.9 h of Indian Supreme Court hearing audio, sentence-aligned by a best-of-both
(MMS-CTC + Whisper) forced-alignment pipeline.

## Results

Whisper `BasicTextNormalizer` applied to both sides; WER/CER via `jiwer`. Trained on the
by-case train split (24,422 clips); both eval sets are held out (the gold hearing is never
trained on). Best checkpoint selected by validation WER.

| Eval set | Model | WER | CER |
|----------|-------|:---:|:---:|
| Held-out test (3,799) | whisper-medium · zero-shot | 12.6 | 7.2 |
| Held-out test (3,799) | **whisper-medium · fine-tuned (this)** | **9.2** | **5.2** |
| Human-verified gold (109) | whisper-medium · zero-shot | 25.0 | 18.1 |
| Human-verified gold (109) | **whisper-medium · fine-tuned (this)** | **14.6** | **10.3** |

Validation WER during training reached **9.53** (CER 5.99). Fine-tuning cuts test WER ~27%
relative and gold WER ~42% relative over zero-shot whisper-medium.

## Training

- **Base:** `openai/whisper-medium` (769M), full fine-tune (no LoRA)
- **Optimizer:** AdamW, lr `1e-5`, 50-step warmup, bf16, gradient checkpointing
- **Batch:** effective 16 (per-device 8 × grad-accum 2)
- **Steps:** 3000 (~2 epochs); decoding forced to English

## Usage

```python
import torch
from transformers import pipeline

asr = pipeline(
    "automatic-speech-recognition",
    model="kirandevraj/whisper-medium-supreme-court-hybrid",
    torch_dtype=torch.float16,
    device=0,  # GPU; use -1 for CPU
)
out = asr("hearing_clip.wav", generate_kwargs={"language": "en", "task": "transcribe"})
print(out["text"])
```

Lower-level (processor + model):

```python
import torch, soundfile as sf
from transformers import WhisperForConditionalGeneration, WhisperProcessor

repo = "kirandevraj/whisper-medium-supreme-court-hybrid"
proc = WhisperProcessor.from_pretrained(repo)
model = WhisperForConditionalGeneration.from_pretrained(repo, torch_dtype=torch.float16).to("cuda").eval()

audio, sr = sf.read("hearing_clip.wav")          # 16 kHz mono
feats = proc(audio, sampling_rate=16000, return_tensors="pt").input_features.to("cuda", torch.float16)
ids = model.generate(feats, language="en", task="transcribe")
print(proc.batch_decode(ids, skip_special_tokens=True)[0])
```

## Notes

- Audio must be **16 kHz mono**; clips ≤ 30 s (the Whisper encoder window).
- Domain: Indian Supreme Court oral hearings — English with occasional romanized Hindi,
  legal vocabulary, case citations, and speaker names.