Views
No views yet
dtype: 'q8', WASM/CPU friendly).openai/whisper-base (73M params) fine-tuned by
Oriserve on ~550 hours of noisy
Indian-accented Hindi. It transcribes Hindi-English code-mixed speech into
romanized Hinglish — English words stay in Latin script.optimum-cli export onnx --task automatic-speech-recognition-with-pastEnableSubgraph for the merged decoder), matching the layout Transformers.js
expects for dtype: 'q8':
onnx/encoder_model_quantized.onnx (23 MB)onnx/decoder_model_merged_quantized.onnx (79 MB)1import { pipeline } from '@huggingface/transformers';
2
3const asr = await pipeline(
4 'automatic-speech-recognition',
5 'TechHirecentive/Whisper-Hindi2Hinglish-Swift-ONNX',
6 { dtype: 'q8' },
7);
8const { text } = await asr(audioFloat32Array16kHz, {
9 language: 'hi',
10 task: 'transcribe',
11});