Synthalingua Version 1.2.5 Change Log

[cyberofficial]

Synthalingua Version 1.2.5 Change Log

Description of Changes

This update introduces a major new feature: an intelligent adaptive batch processing system that dynamically allocates transcription jobs between the GPU and CPU for maximum performance. A comprehensive bug report generator and an SRT subtitle fixer have also been added to improve troubleshooting and usability. Additionally, the setup process for vocal isolation has been significantly simplified by replacing Miniconda with a more lightweight, portable Python environment, and the packaged CUDA version has been updated.

Noticeable Changes

• Intelligent Adaptive Batch Processing: A new --adaptive_batch mode for caption generation intelligently distributes work between your GPU and CPU. It learns your system's performance and allocates longer tasks to the GPU and shorter ones to the CPU, significantly speeding up processing for large files.
• Comprehensive Bug Report Generator: A new --bugreport command has been added. It generates a detailed bugreportinfo.txt file with system, Python, and hardware information to make troubleshooting and reporting issues much easier.
• SRT Subtitle Fixer: A new --fixsrt utility has been added to repair SRT files with out-of-order timestamps, ensuring subtitles are always displayed chronologically.
• Simplified Vocal Isolation Setup: The --using_vocal_isolation feature no longer uses Miniconda. It now sets up a portable, embedded Python environment, reducing the installation footprint, complexity, and potential for environment conflicts.
• Cross-Platform Setup Script: The set_up_env.py script is now fully cross-platform, supporting Windows, Linux, and macOS. Linux users are given a choice between using their system Python or the new embedded Python for vocal isolation.
• Interactive Setup Experience: The setup script is more interactive, guiding users through choices for using existing tools (like FFmpeg, yt-dlp) or downloading fresh copies.
• New About Screen: The --about command now displays a modern, animated interface with project details, features, and contributor acknowledgements.
• HTTPS Server Option: A new --https command-line argument allows running a secure web server alongside the existing HTTP one.
• Redesigned Player Page: The player.html interface has been completely overhauled with a modern design, a two-column layout for settings, and the ability to change the video source directly from the UI. If no video source is provided in the URL, the page now interactively prompts the user for one.
• Enhanced Device Listing: The --list_microphones command now provides a much more detailed and user-friendly table with device ID, channels, and sample rate, along with clear instructions to use the ID for selection.
• Enhanced Remote Microphone Server: The remote_microphone.py script now supports binding the server to a specific IP address using the --serverip argument, allowing for network access beyond localhost.

Hidden Changes

• Adaptive Batch Processing Backend: A new modules/adaptive_batch.py module has been added to handle the logic for GPU/CPU detection, performance tracking, and dynamic job scheduling.
• Web Server Auto-Port Selection: The web server will now automatically search for and use an available port if the one specified is already in use.
• Web Server Security: The Flask web server now includes security enhancements, such as IP blocking after multiple failed requests for static files and improved protection against path traversal attacks.
• Robust Demucs Path Finding: The demucs_path_helper.py script has been rewritten to be more intelligent, actively searching multiple common locations (embedded Python, virtual environments, system paths) to find a valid demucs installation.
• Interactive Timeout Handling: When using --batchmode, if a transcription job times out, the system now interactively prompts the user to either retry with an increased timeout or skip the problematic segment.
• Build Script Simplification: The main build.bat script has been simplified to directly use pyinstaller with the corresponding .spec files, streamlining the build process.
• Discord Message Formatting: The icons in Discord notifications have been replaced with text-based labels (e.g., [ORIGINAL], [TRANSLATION]) for better clarity.

Technical Changes

• Version Update: The local version in modules/version_checker.py has been updated to 1.2.5.
• Argument Parser Update: modules/parser_args.py was updated to include the new --bugreport, --fixsrt, --https, --adaptive_batch, --batchjobsize, --cpu_batches, --max_cpu_time, and --stop_cpu_at arguments.
• New Modules: Added modules/adaptive_batch.py for the new processing scheduler, modules/bug_report.py for generating system reports, and modules/srt_fix.py for the subtitle repair utility.
• Major set_up_env.py Overhaul: The script was rewritten to manage a portable, embedded Python installation instead of Miniconda, handle platform-specific downloads, and generate a bugreportinfo.txt file upon completion.
• Demucs Environment Fix: The set_up_env.py script now sets environment variables (TORIO_USE_FFMPEG=0) to force torchaudio to use the soundfile backend, preventing conflicts with FFmpeg/torchcodec when running demucs.
• CUDA Version Update: The setup scripts (setup.bat, setup.sh) have been updated to use the PyTorch index URL for CUDA 12.9 (cu129) instead of 12.8.
• Transcription Worker Enhancements: The transcribe_worker.py script now accepts a --debug flag for more verbose logging and has improved UTF-8 handling for stability.
• Build Specification Updates: The synthalingua.spec file now includes torchcodec in its data collection and changes the final output directory name to release.
• Transcription Core Stability: The stop_queue_processing method in modules/transcription_core.py was improved to be more robust and prevent race conditions on shutdown.
• Project Organization: Utility scripts have been moved into a misc/ directory, and a new plan.md file has been added, outlining the design of the adaptive batching system.