What should I pay attention to when creating and editing subtitles?
Speech recognition models have difficulty transcribing in situations such as background hum, long silences, or music louder than the speech. In such cases, timestamps may drift. Paying attention to the points below will improve the quality of your subtitle.
• It is a good idea to select 'Powerful and detailed (WhisperX)' under the 'Transcription Engine' option. WhisperX measures the start and end moments of each word at millisecond precision. Although this measurement increases CPU/GPU usage, it brings subtitle synchronization to its best level.
• If you have CUDA support, we recommend using the 'Large' model. If you do not have CUDA support, you can choose the 'Medium' model.
• Running your subtitle file through an AI application for a spelling check will improve the quality of your file. Confusion that occurs, especially in files containing proper names or words from more than one language, can be overcome this way. In MurCr, you can easily copy the AI package by pressing the 'Subtitle' button or using the ALT+SHIFT+C shortcut. For the project instructions you need to give the AI, see this link: How can I fix spelling errors in the subtitle file?
Related topics:
How do I add subtitles?
How do I burn subtitles into a video?
What are the subtitle burn-in options and what do they mean?