huanshere/videolingo

VideoLingo is an automated video translation and dubbing tool written in Python. It takes a video, either a file or a YouTube URL, and produces professionally translated subtitles and optionally a dubbed audio track in another language, aiming for the quality standard used by Netflix subtitle teams. The workflow runs end-to-end without manual intervention: it downloads the video (via yt-dlp for YouTube), transcribes the speech into word-level subtitles using WhisperX (a speech recognition model), segments those subtitles intelligently using NLP (natural language processing, software that understands sentence structure), translates the result using an AI language model, and optionally generates a dubbed audio track using AI voice synthesis. Translation uses a three-step process (translate, reflect, adapt) to handle natural phrasing and cultural context rather than literal word-for-word substitution. A key design choice is that subtitles are always single-line, the README notes this as a deliberate difference from many similar tools, which matches professional broadcast standards. Dubbing supports multiple text-to-speech providers and even voice cloning so the dubbed audio can match the original speaker's voice. The tool runs as a browser-based interface using Streamlit, so you interact with it through a web page rather than a command line. It supports multiple input languages and can translate into any language, though dubbing options depend on which voice synthesis provider is used. You would use VideoLingo to localize YouTube videos, lectures, or any video content for a different language audience without hiring a translation team.