Search Results for "radeon-project"

pyVideoTrans is an ambitious open-source multimedia processing project that assembles speech recognition, subtitle generation, AI translation, voice synthesis, and video assembly into a unified pipeline for converting videos from one language to another with embedded dubbing and captions. At its core it runs speech-to-text models to transcribe audio tracks, translates the resulting text into a target language using local or cloud-based translation engines, synthesizes new speech to match the translated subtitles, and then merges that speech back into the video, creating a fully localized media file. ...

...It is designed to process formats such as EPUB, TXT, SRT, and PDF, then generate translated output that helps readers compare the original text with the target language. The project supports multiple AI providers and models, including OpenAI-compatible models and other translation backends through LiteLLM-style integrations. It is especially useful for public domain books, language learning, subtitle translation, and personal reading workflows. Users can run it from Python scripts or install it as a command-line package for repeated translation tasks. ...

...It transcribes the original speech, translates the text, generates natural-sounding speech in the target language, and remuxes the new audio back into the video. The project is designed to keep the generated speech aligned with the original timing so the final result feels closer to a real dubbed video. It can be used from the command line, through a FastAPI web app, or as a Claude Code skill. Violin supports multilingual workflows and is useful for creators, educators, localization teams, and developers building automated video translation pipelines. ...

...This structure makes the system more steerable than a traditional translation pipeline. Users can adjust prompts to control tone, formality, terminology, idiom handling, and regional language choices. The project includes example scripts and a simple Python interface for translating between source and target languages. It is useful for researchers, developers, and localization teams exploring how LLM-based workflows can produce better, more customizable translations.

...Training relies on denoising auto-encoding and back-translation, with on-the-fly, multithreaded generation of synthetic parallel data to continually refresh supervision signals. The project also provides scripts to fetch and preprocess monolingual data, learn BPE codes, and train cross-lingual embeddings that bootstrap unsupervised alignment between languages. Beyond the core EMNLP 2018 setup, the codebase exposes additional, optional capabilities such as multi-language training, language model pretraining with shared parameters, and adversarial training.