1.4 SRTAITranslate

Release Notes

New Features and Enhancements

Updated Features

1. Model Consistency

   • Maintained the use of the facebook/nllb-200-3.3B model for high-quality translations.

2. Hardware Support

   • Improved support for CUDA-enabled GPUs, with automatic device detection and optimization.

3. Multi-Language Support

   • Retained broad language pair support with consistent accuracy.

4. Implementation of accelerate Module

   • Using this module made the translations go 10 times faster than before.

Refined Functions

1. get_translator(src_lang, tgt_lang)

   • Implemented more robust error handling and logging for model loading issues.
   • Added support for automatic device selection (CUDA or CPU) based on availability.

2. translate_texts(texts, translator, tokenizer, device)

   • Enhanced batch processing with improved error handling and logging.
   • Implemented text cleaning function to improve translation quality.

3. translate_srt(input_file, src_lang, tgt_lang, progress_callback)

   • Optimized batch processing for more efficient SRT file translation.
   • Added support for translation cancellation mid-process.

User Interface Improvements

1. Progress Tracking

   • Implemented a more responsive progress bar with real-time updates.
   • Added a status label to provide more detailed progress information.

2. Error Handling

   • Introduced custom exception classes (TokenInvalidoError, ErroDeTraducao) for more specific error handling.
   • Improved error messages for clearer user feedback during translation and model loading processes.

3. API Token Management

   • Implemented secure token storage using the keyring library.
   • Added a toggle button to show/hide the API token for enhanced security.

4. Cancellation Feature

   • Added a "Cancel" button to allow users to stop the translation process.

5. UI Layout and Styling

   • Improved overall layout and styling of the application for better user experience.
   • Implemented consistent styling using ttk widgets and custom styles.

New Additions

1. Multithreading

   • Implemented multithreading for the translation process to prevent UI freezing during long operations.

2. Logging

   • Added comprehensive logging throughout the application for better debugging and issue tracking.

Bug Fixes

1. Translation Consistency

   • Addressed potential issues in token generation and decoding for more reliable translations.
   • Implemented text cleaning to remove problematic characters and improve translation quality.

2. UI Responsiveness

   • Improved threading implementation to enhance UI responsiveness during translation tasks.

3. Memory Management

   • Optimized memory usage during translation to handle larger SRT files more efficiently.

Instructions for Use

• Ensure all required dependencies are installed using Module Installation.py.
• Run the main script SRTAI_Translator.py to launch the application.
• Use a valid Hugging Face API token for accessing the translation model.

If you are using the .EXE version

• Download both binaries
• Merge them extracting both together

Thank you for using SRT AI Translator!

Full Changelog: 1.3...1.4

1.3 SRTAITranslate

Release Note

New Features and Enhancements

Updated Features

1. Model Consistency

   • Maintained the use of the facebook/nllb-200-3.3B model for high-quality translations.

2. Hardware Support

   • Continued support for NVIDIA GPUs (CUDA), Apple Silicon (MPS), and AMD GPUs (ROCm) for optimized translation performance.

3. Device Detection

   • Retained automatic detection of CUDA, MPS, and ROCm-compatible devices for best performance.

4. Multi-Language Support

   • Kept broad language pair support with consistent accuracy.

Refined Functions

1. get_translator(src_lang, tgt_lang)

   • Improved error handling and messaging for model loading issues.

2. translate_texts(texts, translator, tokenizer, device)

   • Enhanced logging for better tracking of batch translations.

3. translate_srt(input_file, src_lang, tgt_lang, progress_callback)

   • Optimized batch processing for more efficient SRT file translation.

User Interface Improvements

1. Progress Tracking

   • Refined progress bar update mechanism for smoother visual feedback.

2. Error Handling

   • Further improved error messages for clearer user feedback during translation and model loading processes.

3. API Token Management

   • Streamlined token handling process with improved security measures.

New Addition

1. Installation Script
   • Added a new Python script to automate the installation of required modules, including PyTorch and other dependencies.

Bug Fixes

1. Translation Consistency

   • Addressed potential issues in token generation and decoding for more reliable translations.

2. UI Responsiveness

   • Improved threading implementation to enhance UI responsiveness during translation tasks.

Instructions for Use

• For the executable version: Extract both parts of the zip archive to merge them into the .exe file.
• For the Python script version: Run the new installation script to set up all required dependencies before using the main application.

Thank you for using SRT AI Translator!

1.2 SRTAITranslate

Release Note

New Features and Enhancements

Updated Features

1. Hardware Support Expansion

   • Added support for NVIDIA GPUs using CUDA and AMD/Intel GPUs through ROCm for enhanced translation performance.

2. Model Enhancement

   • Switched from using the MarianMTModel to the facebook/nllb-200-3.3B model for improved translation quality.

3. Device Detection

   • Implemented automatic detection of CUDA and ROCm-compatible devices for optimized performance.

4. Multi-Language Support

   • Updated language support to cover more language pairs with enhanced accuracy.

New Functions

1. get_translator(src_lang, tgt_lang)

   • Returns a translator function, tokenizer, and model based on the source and target languages.
   • Now supports both CUDA and ROCm devices.

2. translate_texts(texts, translator, tokenizer, device)

   • Translates a batch of texts using the provided translator function and tokenizer.
   • Added functionality to handle device-specific operations for CUDA and ROCm.

3. translate_srt(input_file, src_lang, tgt_lang, progress_callback)

   • Reads and translates an SRT file, saving the translated output to a new file.
   • Supports CUDA and ROCm for efficient processing.

User Interface Improvements

1. Progress Bar

   • Added a progress bar to indicate translation progress.

2. Error Handling

   • Enhanced error messages for better user feedback on issues during translation and model loading.

3. API Token Management

   • Improved token handling by saving it to a configuration file and reloading it as needed.

Bug Fixes

1. Translation Errors

   • Fixed issues related to token generation and decoding for more accurate translations.

2. Device Compatibility

   • Resolved compatibility issues with different hardware configurations.

Instructions to use the .EXE

• Just extract both parts of the zip archive, and they will merge into the .exe.

Thanks for using our software!

1.1 SRTAITranslate

Added the funcionality that automatically saves the last used Huggingface API Key.

Any comments or suggestions are welcome :D

What's Changed

• README.md by @Kattiell in #3
• exe update by @MaloneFreak
• README.md by @Kattiell in #5

Full Changelog: https://github.com/MaloneFreak/SRTAITranslate/commits/1.1

1.0 SRTAITranslate

Full Changelog: https://github.com/MaloneFreak/SRTAITranslate/commits/release