Skip to content

CHANGELOG.md

Latest commit

 

History

History
361 lines (217 loc) · 9.91 KB

CHANGELOG.md

File metadata and controls

361 lines (217 loc) · 9.91 KB
Raw

1.7.0a1

Add a token-efficient lean translation pipeline, broaden model-provider routing, and refresh the configuration surface for mixed-model workflows.

New Features:

  • Add LeanTranslator with compact anchored prompts, sliding context, checkpoint resume, retry handling, binary-split recovery, and atomic fallback for missing lines.
  • Add lean mode integration to LRCer through TranslationConfig(translate_mode="lean"), including optional Context Review and separate CR/translation chatbots.
  • Add LiteLLM support as an optional provider gateway via openlrc[litellm] and litellm:<provider/model> routing.
  • Add ModelConfig.extra_body passthrough for provider-specific request parameters across OpenAI-compatible, Anthropic, Gemini, and LiteLLM backends.

Other Changes:

  • Replace legacy string-based translation config fields with ModelConfig objects for primary, retry, and Context Review chatbots.
  • Decouple model capability limits from the static model registry; max_tokens and context_window are now opt-in runtime caps.
  • Improve custom-provider and provider-prefix inference for OpenAI-compatible, Anthropic, Gemini, LiteLLM, and third-party model names.
  • Restrict lingua language detection to the supported language set and add anchor-based validation for lean translation output.
  • Split media helpers from utils into media_utils to keep heavyweight media dependencies isolated.
  • Centralize live API test configuration, add LeanTranslator and LiteLLM regression coverage, and update CI to exercise the new optional litellm extra.
  • Refresh README examples for lean mode, mixed-model translation, live test configuration, and the current local-LLM/provider support status.

1.6.3

Improve translation reliability, reduce heavy import overhead, and tighten release tooling around the uv workflow.

New Features:

  • Add token-aware, scene-sensitive chunk splitting for long subtitle translation.
  • Add chunked guideline generation for long subtitle files.

Other Changes:

  • Fully defer heavyweight package-root imports and move torch/deepfilternet behind the optional full extra.
  • Harden ChatBot and GeminiBot lifecycle handling, retry behavior, and error recovery.
  • Reduce translation log noise and add OPENLRC_LOG_LEVEL.
  • Improve line-mismatch retry behavior and raise the default context window to 32K.
  • Replace runtime assert checks with explicit exceptions.
  • Fix empty-translation early return, continuous-script duplicate word handling, and lazy import test stdout parsing.
  • Gate long stress tests behind OPENLRC_TEST_STRESS and expand chunking/translation coverage.

1.6.2

Refine the transcription and translation pipeline, expand test coverage, and improve the release workflow around uv.

New Features:

  • Add TranscriptionConfig and TranslationConfig dataclasses for clearer runtime configuration.
  • Decouple transcription and translation in LRCer.
  • Add OpenRouter fallback model support.

Other Changes:

  • Add skip_preprocess support.
  • Fix retry_model validation.
  • Improve duplicate handling and video format inference in the decoupled flow.
  • Refresh README examples to the new config API.
  • Add mock and live regression coverage for translator and LRCer behavior.
  • Add Ruff and Pyright checks and update CI environment variables for live API tests.

1.6.1

Replace poetry with uv package manager. Fix #69.

Other Changes:

  • Hot fix for missed mp3 files issue.
  • Allow un-recorded (not in models.py) model name.

1.6.0

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.

1.5.2

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.

1.5.1

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.

1.5.0

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.

1.4.1

Minor bugfixes and improvements.

Other Changes:

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

1.4.0

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.

1.3.1

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.

1.3.0

Add basic GUI support via streamlit.

Other Changes:

  • Add clear_temp_folder args.

1.2.0

This update add Claude Models support for translation.

1.1.0

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.

1.0.3

This update fix minor issues.

Other Changes:

  • Remove water mark in srt and lrc.
  • Improve logging.

1.0.2

This update add minor features.

Other Changes:

  • Binlingual subtitle support (Beta).
  • Improve translation prompt.

1.0.1

This update fix minor issues.

Other Changes:

  • Fix issue that prevent the usage of whisper-large-v3.
  • Remove tags from translation texts.

1.0.0

This update introduce new features.

New features:

  • Resume from previous translation.
  • Add atomic translation for src-trans inconsistency.

Other Changes:

  • Update default whisper model to whisper-large-v3.

0.2.3

This update improves preprocess efficiency and minor changes.

Other Changes:

  • Introduce multiprocessing for loudness normalization.
  • Fix the .srt generation issue for video input.
  • Add preprocess options, which users can tune.

0.2.2

This update addresses minor issues.

Other Changes:

  • Split audio during noise suppression to avoid out-of-memory.
  • Improve translation prompt.

0.2.1

This update adds a preprocessor to enhance input audio (loudness normalization & noise suppression).

New features:

Other Changes:

  • Now all the intermediate files are saved in ./path/to/audio/preprocess.

0.2.0

This update switch the underlying transcription model from whisperx to faster-whisper, which enable VAD parameter tuning.

New features:

  • Switch whisperx back to faster-whisper for VAD parameter tuning.

Other Changes:

0.1.5

Emergent bugfix release.

0.1.4

This update add input video support and introduce context configuration.

New features:

  • Add word_align and sentence_split for non-word-boundary languages to split long text into sentences.
  • Add text-normalization for help matching sentences.
  • Add skip-translation support.

Other Changes:

  • Use pathlib to handle paths.
  • Improve timeline accuracy.

0.1.3

This update add input video support and introduce context configuration.

New features:

  • Add input video support.
  • Add context configuration for inputs.

Other Changes:

  • Add test suites to CI.
  • Add language detection for translated content.
  • Improve prompt by adding background info.
  • Update punctuator model.
  • Replace opencc with more light-weight zhconv.

0.1.2

This update improves the timeline consistency of translated subtitles. Thanks gpt-subtrans!

New features:

  • Fix misaligned timeline issue by improving translation prompt.
  • Add output srt format support.
  • Add changeable temperature and top_p parameter for GPTBot.

Other Changes:

  • Report total OpenAI translation fee for multiple audios.
  • Improve repeat-checking algorithm.

0.1.1

This update enhances the efficiency of processing multiple audio files.

New features:

  • Implementation of a producer-consumer model to process multiple audio files.

Other Changes:

  • Update logger with colored format.
  • Minor parameter modification that makes the timeline of translation more intuitive.

0.1.0

This update significantly improves translation quality, but at the cost of slower translation speed.

New Features:

  • Use multi-step prompt for translation.
  • Update the default model to gpt-3.5-turbo-16k.
  • Automatically fix json encoder error using GPT.

Other Changes:

  • Calculate the accurate price for OpenAI API requests.

0.0.6

This update greatly improves the quality of transcription (both in time-alignment and text-quality).

New Features:

  • Use whisperx to improve transcription accuracy.
  • Add Traditional Chinese to Mandarin optimization when target_lang=zh-cn.

0.0.5

New Features:

Other Changes:

  • Update build tool to poetry.

0.0.4

New Features:

  • Use async call to communicate with OpenAI api.
  • Abstract the GPT communication module as GPTBot.
  • Add fee limit for GPTBot.