Whisper Hindi Apex — MLX
MLX-converted weights for Oriserve/Whisper-Hindi2Hinglish-Apex.
- Architecture: whisper-large-v3-turbo (809M params, 4 decoder layers)
- Fine-tuned on: 700+ hours of Hindi/Hinglish audio
- Output: Romanized Hindi (Hinglish)
- Format: weights.npz (MLX Whisper compatible)
Usage
import mlx_whisper
result = mlx_whisper.transcribe(
"audio.wav",
path_or_hf_repo="knownsense/whisper-hindi-apex-mlx",
language="hi",
word_timestamps=True,
)
Converted from Oriserve/Whisper-Hindi2Hinglish-Apex using convert_apex_to_mlx.py.
- Downloads last month
- 338
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for knownsense/whisper-hindi-apex-mlx
Base model
openai/whisper-large-v3 Finetuned
openai/whisper-large-v3-turbo Finetuned
Oriserve/Whisper-Hindi2Hinglish-Apex