Major Update: This release introduces intelligent adaptive batch processing that dynamically balances work between your GPU and CPU for maximum performance. I've also simplified vocal isolation setup, added powerful new utilities, and made numerous improvements across the board.

Major New Features

Intelligent Adaptive Batch Processing

The new --adaptive_batch mode learns your system's performance and intelligently distributes transcription jobs between GPU and CPU. Longer tasks go to the GPU while shorter ones leverage your CPU, significantly speeding up processing for large files.

Bug Report Generator

Use --bugreport to generate a comprehensive bugreportinfo.txt file with detailed system, Python, and hardware information. Makes troubleshooting and issue reporting much easier!

SRT Subtitle Fixer

The new --fixsrt utility repairs SRT files with out-of-order timestamps, ensuring your subtitles always display chronologically.

Simplified Vocal Isolation Setup

Vocal isolation setup has been completely overhauled. We've replaced Miniconda with a lightweight, portable Python environment that reduces installation footprint, complexity, and potential conflicts.

User Experience Improvements

• Cross-Platform Setup: The setup script now fully supports Windows, Linux, and macOS with an interactive experience
• Modern About Screen: New --about command displays an animated interface with project details and contributor acknowledgements
• HTTPS Server: Run a secure web server with the new --https argument
• Redesigned Player: Complete overhaul of player.html with modern design, two-column settings layout, and interactive video source selection
• Enhanced Device Listing: --list_microphones now shows detailed tables with device ID, channels, and sample rate
• Remote Microphone Server: Bind to specific IP addresses using --serverip for network access

Technical Improvements

• CUDA 12.9 Support: Updated to latest CUDA version for better GPU performance
• Auto-Port Selection: Web server automatically finds available ports
• Enhanced Security: IP blocking after failed requests and improved path traversal protection
• Robust Demucs Detection: Smarter path finding across embedded Python, virtual environments, and system paths
• Interactive Timeout Handling: Batch mode now prompts to retry or skip problematic segments
• Improved Discord Notifications: Text-based labels for better clarity
• Stability Fixes: Better UTF-8 handling and race condition prevention

New Command-Line Arguments

• --adaptive_batch - Enable intelligent GPU/CPU job distribution
• --batchjobsize - Control adaptive batch job sizing
• --cpu_batches - Specify CPU batch configuration
• --max_cpu_time - Set maximum CPU processing time
• --stop_cpu_at - Define CPU processing cutoff
• --bugreport - Generate system diagnostic report
• --fixsrt - Repair subtitle timestamp ordering
• --https - Enable secure web server

Under the Hood

This release includes significant architectural improvements:

• New modules/adaptive_batch.py for intelligent job scheduling
• New modules/bug_report.py for system diagnostics
• New modules/srt_fix.py for subtitle repair
• Complete set_up_env.py rewrite for portable Python management
• Environment variables set to prevent FFmpeg/torchcodec conflicts
• Improved build process with simplified scripts
• Better project organization with misc/ directory for utilities

Thank you for using Synthalingua! I hope these improvements enhance your transcription workflow.