Model Usage Recommendations

Do not use Large-v2, Large-v3, and Large-v3-Turbo models on low-performance computers as they require high device performance, which may cause lag or unresponsiveness.

Different models are recommended for different scenarios. Here are some suggestions.

File Transcription

• Low-performance devices: Recommended to use Tiny and Base from the official Whisper models. These smaller models ensure transcription speed on low-performance devices.
• Medium-performance devices: Recommended to use Small and Medium models, which balance speed and accuracy.
• High-performance devices: Recommended to use Large-v2, Large-v3, and Large-v3-Turbo models for the highest transcription accuracy.

Real-time Transcription

• Low-performance devices: Recommended to use the Real-time Model, specifically designed for real-time scenarios with low device performance requirements.
• Medium-performance devices: Recommended to use Small and Medium from the official Whisper models, providing good results in real-time transcription.
• High-performance devices: Recommended to use Large-v2, Large-v3, and Large-v3-Turbo from the official Whisper models for optimal real-time transcription accuracy.

Summary in table format:

Users with high-performance devices are advised to directly use the Large-v3-Turbo model, as it maintains accuracy comparable to Large-v3 while improving inference speed.

Model-related Discussions

• Most accurate model for English?