faster-whisper-v2-d4
| Property | Value |
|---|---|
| Base Model | openai/whisper-large-v2 |
| License | Apache 2.0 |
| Languages | Hebrew, English |
| Training Data | 350 hours (250h volunteer + 100h professional) |
What is faster-whisper-v2-d4?
faster-whisper-v2-d4 is an optimized speech recognition model developed by ivrit-ai, specifically designed for Hebrew and English language processing. Built upon OpenAI's Whisper large-v2 architecture, this model leverages the faster-whisper implementation for improved performance and efficiency.
Implementation Details
The model was trained on a diverse dataset comprising 250 hours of volunteer-transcribed speech from the ivrit-ai/crowd-transcribe-v4 dataset and an additional 100 hours of professionally transcribed speech. This combination ensures both breadth and quality in the training data.
- Optimized for faster inference using the faster-whisper framework
- Built on the robust Whisper large-v2 architecture
- Specialized for Hebrew and English language processing
Core Capabilities
- Accurate transcription of Hebrew and English speech
- Easy integration through the faster_whisper Python package
- Support for language-specific transcription options
- Efficient processing and improved performance
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized optimization for Hebrew language processing, combined with the efficiency benefits of the faster-whisper implementation. The diverse training dataset, including both crowd-sourced and professional transcriptions, provides robust performance across various speech contexts.
Q: What are the recommended use cases?
The model is ideal for applications requiring Hebrew or English speech transcription, particularly in scenarios where processing efficiency is crucial. It's suitable for both production environments and research applications, thanks to its Apache 2.0 license.
