subvid.app is a subtitle tool that runs entirely inside your web browser. You drop in a video file, and it generates subtitles automatically using an AI speech recognition model called Whisper. If you need subtitles in a different language than the original audio, it can translate them too, using a separate translation model. No account, no uploads to a server, and no API keys are required. Your video file stays on your device the whole time. Once subtitles are generated, you can edit them in a built-in timeline editor. You can fix the text, adjust timing, change fonts, colors, and position, and manage multiple language tracks. When you are happy with the results, you can export either a plain .srt subtitle file or a new video with the captions burned directly into the picture. The AI models (Whisper for speech and NLLB-200 for translation) are downloaded from Hugging Face the first time you use them and then cached locally in your browser. They are roughly 150 MB combined. After that first download the tool works offline. On browsers that support WebGPU, the transcription step runs faster. The project is built with the Astro web framework, styled with Tailwind CSS, and deployed as static files on Cloudflare Workers. It supports English and Spanish interfaces. Developers can clone the repo, run pnpm install and pnpm dev, and have a local version running immediately without any environment setup. The codebase separates heavy AI work into background Web Workers so the editing interface stays responsive while models are running.
← midudev on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.