v1.6.0

Update faster-whisper to the latest version. Add models.py for model info.

Other Changes:

• Fixed issue #60.
• Change default parameters for new faster-whisper.
• Update installation guide in README.

v1.5.2

Code refactoring, documentation updates, and minor bug fixes.

Other Changes:

• Fixed issue #54.
• Changed the default LLM model to gpt-4o-mini.
• Minor improvements on prompts.
• Improve post-optimizations for subtitles.

v1.5.1

Minor bugfixes and improvements.

Other Changes:

• Fix default check_format returning False issue.
• Fix edges cases for srt file generation.
• Prepare translation evaluator for benchmarking.

v1.5.0

This update add Gemini Models support for translation.

New Features:

• Support Gemini as translation engine.

Other Changes:

• Extract the check_format section into a standalone validators.py module.
• Add stop_sequences support for Chatbot.
• Use stop sequences when building Context.
• Also remove generated .wav files from videos if clear_temp=True.
• Minor improvements on Context Reviewer prompt.

v1.4.1

Minor bug fixes and improvements.

Other Changes:

• Improve timestamp accuracy.
• Fix edges cases for transcription.

v1.4.0

New Features:

• Add glossary support for domain specific translation.
• Add retry_model args for retrying translation with different models.
• Introduce privider: model_name for arbitrary model routing.

Other Changes:

• Extend 0.5s for each suitable sentence.
• Reduce noise suppression chunk-size for faster processing.
• Enhance translation workflow by Context Reviewer Agent.

v1.3.1

New Features:

• Add custom endpoint (base_url) support for OpenAI & Anthropic.
• Generating bilingual subtitles.

Other Changes:

• Fix dep issues from ctranslate2 and streamlit-related packages.

v1.3.0

Add basic GUI support via streamlit.

Other Changes:

• Add clear_temp_folder args.

v1.2.0

This update add Claude Models support for translation.

v1.1.0

This update improve the translation quality.

Other Changes:

• Update faster-whisper version to 1.0.0.
• Set hallucination_silence_threshold to 2, which alleviates the hallucination issue.
• Add proxy argument.