Fix for loaded model generation

Fixed the need of speaker_wav when a model is loaded speaker_wav is optional in the api call because by default there's reference.wav in a fine trained model and so I modified the code to look for this reference.wav when there's no valid speaker_wav reference. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit - **Chores** - Expanded `.gitignore` to prevent tracking of unnecessary directories and files. - **Improvements** - Updated the app version default value to "main" for better clarity. - **New Features** - Introduced functionality to switch TTS models and send requests to a local server. - Enhanced TTS functionality with support for optional speaker voices and language selection. - **Bug Fixes** - Fixed import paths to ensure proper module organization. <!-- end of auto-generated comment: release notes by coderabbit.ai -->