whisperkit-hinglish

WhisperKit Hinglish (CoreML)

CoreML models converted from Oriserve/Whisper-Hindi2Hinglish-Apex for use with WhisperKit on Apple devices.

Model Details

What This Model Does

Transcribes Hindi and Hinglish speech directly to romanized Latin text — no transliteration needed.

Example output: "yaar mai kal office nahi jaunga" (not "यार मैं कल ऑफिस नहीं जाऊँगा")

Usage with WhisperKit (Swift)

import WhisperKit

let config = WhisperKitConfig(
    model: "Oriserve_Whisper-Hindi2Hinglish-Apex",
    modelRepo: "shrimalmadhur/whisperkit-hinglish"
)
let whisperKit = try await WhisperKit(config)
let result = try await whisperKit.transcribe(audioPath: "audio.wav")
print(result.text)

Important: Use language: "en" in decoding options — the model internally maps Hindi speech to English/Latin tokens.

Contents

• Oriserve_Whisper-Hindi2Hinglish-Apex/AudioEncoder.mlmodelc
• Oriserve_Whisper-Hindi2Hinglish-Apex/TextDecoder.mlmodelc
• Oriserve_Whisper-Hindi2Hinglish-Apex/MelSpectrogram.mlmodelc
• Oriserve_Whisper-Hindi2Hinglish-Apex/config.json
• Oriserve_Whisper-Hindi2Hinglish-Apex/generation_config.json

Credits

• Source model: Oriserve/Whisper-Hindi2Hinglish-Apex by OriserveAI
• Conversion tools: whisperkittools by Argmax