Update Synthalingua Version 4.63

[cyberofficial]

Synthalingua Changelog

Portable + GUI is at itch.io! Pick it up there.

-- Please allow some time for itch.io to process the update --

[1.1.0] - 2025-02-21

This release focuses on significant code restructuring, UI enhancements, and dependency updates for improved performance, stability, and user experience. It also addresses several bugs and incorporates community contributions.

Added

• Customizable Word Block List: Implemented the --ignorelist flag to load a text file containing words or phrases to be ignored from the output.
• Improved Caption Customization in Player: Enhanced caption customization options (background color, text color, font size, position, opacity, width) in the web browser player.
• Support for Password-Protected HLS Streams: Added --remote_hls_password_id and --remote_hls_password flags.
• Support for 7gb Turbo Model: Added support for the 7GB turbo model, optimized for faster transcription with minimal accuracy loss.
• Dependency Updates: Updated core dependencies (including yt-dlp, openai-whisper) for performance and feature improvements.
• User Mode Selection: Added a User Mode selection screen to the GUI, allowing users to choose between Basic and Advanced modes.
• Ko-fi Button in GUI: Integrated a Ko-fi button for donations.

Changed

• Code Restructuring:
  • Refactored modules and classes for better organization and maintainability.
  • Replaced modules/imports.py with direct imports.
  • Created new modules: audio_handlers.py, device_manager.py, file_handlers.py, stream_handler.py, and transcription_core.py.
• Command Line Arguments: Replaced --auto_model_swap with RAM-based model selection.
• Microphone Calibration: Improved with concurrent calibration and listening using separate microphone instances.
• Caption Generation: Updated default parameters and language handling.
• Stream Processing: Enhanced segment downloading, error recovery, and buffer management.
• Temporary File Handling: Implemented cleanup of temporary files.
• Console Output: Improved logging, added --no_log option for simplified output.
• Error Handling: Standardized error reporting and optional Discord webhook notifications.
• Subtitle Generation: Refactored with chunked processing, confidence scoring, and custom model directory support.
• GUI Updates: Improved layout, button accessibility, and tooltips. Reorganized elements, including settings tabs and added a quick save button.
• Renamed 12GB Model to 11GB: Reflects model size changes.
• 7 GB Option: Enabled a compressed 11GB model option for use with 7GB of VRAM, providing similar results with potential minor accuracy differences.
• Configuration and Command Generation: Implemented classes for managing UI configuration saving/loading and command-line argument generation.
• File and Web Management: Added classes to handle file/directory operations and web-related tasks.

Removed

• Deprecated --auto_model_swap Flag: Removed due to improved model selection.
• Fine-Tuned Model Support (Temporary): Disabled for retraining. Removed obsolete files checkenv.py and warnings.py.

Fixed

• DVR Playback Issues: Improved stream handling, playback control, and buffer management for stable DVR playback.
• Logging Issues: Improved formatting, error reporting, and file management for logs.
• Issues --no_log / cant see any transcriptions/translations #149 and Questions regarding usecase - software defined radio/radio comms #150: Attempted fixes for stability and error handling.
• UI Scaling Issues: Fixed window scaling issues for various resolutions.
• Control Responsiveness: Improved GUI control responsiveness.
• Visual Feedback: Enhanced visual feedback for GUI interactions.
• Remote Microphone Script Password Handling: Fixed password handling.
• Torch Installation Error: Clarified Python 64-bit requirement.
• Caption Generation Filename Error: Fixed filename handling for non-English characters.

---

This discussion was created from the release Update Synthalingua Version 4.63.