---
license: apache-2.0
language: [id, ind]
tags: [automatic-speech-recognition, onnx, onnx-asr, whisper, indonesian]
base_model: cahya/whisper-medium-id
---

# whisper-medium-id — ONNX

ONNX export of [cahya/whisper-medium-id](https://huggingface.co/cahya/whisper-medium-id)
(Whisper medium fine-tuned for Indonesian by [cahya](https://huggingface.co/cahya)) for
[onnx-asr](https://github.com/istupakov/onnx-asr) (standard `whisper` model type — works with
stock onnx-asr, no patches needed). fp32 and int8 variants included.

License: apache-2.0, inherited from the source model.

## Usage

```python
import onnx_asr
model = onnx_asr.load_model("whisper", "path/to/this/repo")  # or quantization="int8"
print(model.recognize("audio_16khz.wav", language="id"))
```

Verified on a FLEURS Indonesian (`id_id`) test clip:

- Reference: "tim-tim virtual memiliki standar keunggulan yang sama dengan tim konvensional tetapi ada sedikit perbedaan"
- fp32: "Tim-tim virtual memiliki standar keunggulan yang sama dengan tim konvensional, tetapi ada sedikit perbedaan." (RTF 0.87)
- int8: identical transcript (RTF 0.34)

Both exact matches to the reference (modulo punctuation/casing). RTF measured on an AMD
Ryzen 5 7600 (CPU, 4 OMP threads, shared/loaded box — not a clean benchmark number).

Int8 decoder was produced by quantizing the pre-merge decoders separately and re-merging
(`merge_decoders(..., strict=False)`); direct quantization of the merged decoder graph does
not shrink it (its `If` subgraphs are skipped by onnxruntime's dynamic quantizer).