Mesin Video Otomatis turns a single topic into a short-video production package: an editable storyboard, per-scene narration read aloud in the browser (Web Speech API), a video preview rendered from canvas (MediaRecorder WebM), and a downloadable package (script .md, subtitles .srt, project .json). Everything runs as a static site — no backend, no account, no cost. The Frankenstein stitch: four technologies that usually stand alone, sewn into one workflow: (1) A rule-based text generator — not a paid LLM; deterministic, runs locally, three styles (Educational/Story/Promo). (2) Web Speech API — narration read by the browser's system voice; id-ID when available, default otherwise. (3) Canvas 2D + MediaRecorder — the preview renders from a 1280x720 canvas via captureStream(30) and records as WebM; no render server, no ffmpeg. (4) Style templates + production-package export. Built with: Kiro (spec-driven development — requirements EARS, design, tasks 4 waves 13 tasks, steering, hooks in the .kiro/ folder), HTML5 + CSS3 + JavaScript ES Modules with no framework and no build step, Web Speech API, Canvas 2D + MediaRecorder, GitHub Pages. Honest limitation: the recorded video has no audio (captureStream captures canvas pixels only, not speechSynthesis audio; narration is still heard during the live preview).
Built With
- canvas-2d
- css3
- github
- html5
- javascript
- kiro
- mediarecorder
- web-speech-api
