whisper-small-dv-mlx

This model was converted to MLX format from alakxender/whisper-small-dv-full.

A fine-tuned Whisper Small model for Dhivehi (Thaana) speech recognition, optimized for Apple Silicon via MLX.

Use with mlx-whisper

pip install mlx-whisper

import mlx_whisper

result = mlx_whisper.transcribe(
    "audio.wav",
    path_or_hf_repo="mohamedrayyan/whisper-small-dv-mlx",
)
print(result["text"])

Original Model

• Base model: openai/whisper-small
• Fine-tuned on: alakxender/dv-audio-syn-lg (Dhivehi audio dataset)
• eval_wer: 2.3726
• eval_wer_ortho: 15.8343
• Parameters: ~244M
• Format: float16 MLX safetensors