// the find
YaoFANGUK/video-subtitle-generator
视频音频生成字幕，生成srt文件。无需申请第三方API，本地实现音频转文本。基于Transformer的视频字幕生成框架。A GUI tool for generating subtitle from videos and generating srt files.
A local Whisper-based GUI tool that transcribes audio/video files and outputs SRT subtitle files, with no external API required. It ships pre-built Whisper model weights and a Tkinter GUI, targeting Windows/Mac/Linux users who want offline subtitle generation. The primary audience is Chinese-speaking users wanting to subtitle Chinese-language content, though it supports ~80 languages.
Ships pre-compiled ffmpeg binaries for all three platforms so users don't need to install it separately. Bundles the Whisper model weights directly in the repo across base/medium/large tiers, so setup is genuinely offline after clone. The CLI interface supports all Whisper language codes and is scriptable for batch use. GPU VRAM requirements are clearly documented per model size, which is the most common stumbling block for local ML tools.
Last commit was February 2024 — Whisper has moved significantly since (faster-whisper, whisper.cpp, insanely-fast-whisper all exist now and are substantially faster). Bundling model weights in git is an anti-pattern: the repo is bloated, you can't easily update models, and large_asr has 62 separate model shard files checked in. Requires an Nvidia GPU with no CPU fallback path mentioned, which cuts out a large chunk of potential users. The README is almost entirely in Chinese with no English translation beyond the title, which limits adoption despite the English description claiming otherwise.