wavecap-whisper

What problem does it solve?

This Skill allows users to fine-tune the WaveCap Whisper speech-to-text model for optimal transcription accuracy and performance based on their specific needs and hardware.

Core Features & Use Cases

• Model Selection: Choose from various Whisper model sizes (tiny, base, small, medium, large-v3) and backends (auto, mlx, faster-whisper) to balance speed and accuracy.
• Decoding Parameter Tuning: Adjust beam size, temperature, and conditioning on previous text for finer control over transcription output.
• Prompt Engineering: Configure global or named initial prompts to improve recognition of domain-specific vocabulary and acronyms.
• Use Case: A user experiencing frequent misinterpretations of technical jargon in their audio streams can use this skill to provide a custom prompt and select a more accurate model, significantly improving transcription quality.

Quick Start

Use the wavecap-whisper skill to set the Whisper model to large-v3-turbo with a beam size of 8 and temperature 0.0.