Show HN: Open Video Overview – Generate narrated videos from text with AI (github.com)

Open Video Overview is an experimental open‑source project (Show HN) that converts plain text into narrated, stylized videos via an interactive Mastra playground. It stitches together a storyboard, Gemini‑generated images, ElevenLabs TTS narration, and ffmpeg processing to produce MP4 explainers or short vertical clips. The repo supports 25+ visual styles (anime, watercolor, 3b1b, retro, whiteboard, etc.), 16 languages, selectable aspect ratios (16:9, 9:16), and narrative tones—making it a quick way to prototype educational explainers, repo walkthroughs, or short social videos. It’s MIT‑licensed and inspired by NotebookLM’s Video Overview. Technically, the videoGenerationWorkflow accepts required params (content, style, format, aspectRatio) and optional ones (language, customInstructions, narrativeVoiceId, narrativeStyle). The pipeline generates a storyboard from source content, creates per‑clip transcripts, renders images with Gemini, synthesizes audio with ElevenLabs, combines each clip, then concatenates the final video into output/<project>-<timestamp> with folders for images, audio, and clips. Requirements: Node.js 18+, pnpm, ffmpeg/ffprobe, plus Google Generative AI and ElevenLabs API keys; run locally with pnpm run dev and visit http://localhost:4111. Expect rough edges—this is a rapidly built experiment with PRs welcome.