ZoyClip - Note to Short Video

approved

by zoy

Turn a note into a 9:16 vertical short video. A cloud LLM (OpenAI/DeepSeek) rewrites it into a script; local Kokoro TTS and ffmpeg render it with auto images, captions and music. Your note text is sent to the cloud LLM. - This plugin has not been manually reviewed by Obsidian staff.

↓ 87 downloadsUpdated 1mo agoMIT
Install in ObsidianView on GitHub

ZoyClip — Note → Vertical Short Video

English | 中文

Turn any note into a faceless 9:16 vertical short video for TikTok, Instagram Reels, YouTube Shorts and 小红书 (RED) — with local AI voiceover, auto images, animated captions, background music, and an auto-generated cover. Bilingual (English / 中文). Runs locally, free, no watermark.

One note in → a finished 1080×1920 MP4 (plus a cover image) embedded right back into your note.

What it does

  1. Script — rewrites your note into a punchy spoken-word script (hook + short segments) via a cloud LLM (OpenAI or DeepSeek, OpenAI-compatible).
  2. Voiceover — synthesizes narration locally with Kokoro TTS (via sherpa-onnx), English and Chinese. Engine + model auto-download once, then fully offline — no Python, no server. Falls back to the system voice.
  3. Images — auto-fetches a fitting photo per segment from Pexels (by AI-generated keywords), or uses images embedded in your note / a folder. Adds Ken Burns zoom-pan and cross-fades.
  4. Captions — clean single-line captions that follow the speech (animated, big, readable; CJK-aware).
  5. Music — optional background music with automatic ducking (music dips under the voice).
  6. Cover — auto-generates a cover image (title + first image) for the thumbnail / click-through.
  7. Render — composites everything with ffmpeg into a 9:16 MP4 and embeds it in your note.

Language

Switch the whole plugin — the UI and the generated video (script, captions, voiceover) — between English and 中文 in settings. Default Auto follows your Obsidian UI language.

Privacy & network use

ZoyClip is local-first. Every network call below is triggered by you when you produce a video — none for analytics or telemetry:

  • Cloud LLM (required). Your note's text is sent to the LLM endpoint you configure (OpenAI api.openai.com, DeepSeek api.deepseek.com, or any OpenAI-compatible URL you enter) using your own API key, to rewrite it into a script. This is the only step that sends your note content off your machine.
  • Pexels (optional). If you add a Pexels API key, short English image keywords (not your note text) are sent to api.pexels.com to fetch stock photos.
  • One-time downloads of open-source tools from public GitHub Releases, then run locally via Node child_process:
    • ffmpeg static binary — from eugeneware/ffmpeg-static (macOS only; or use your own ffmpeg via the path setting).
    • sherpa-onnx TTS engine + Kokoro voice models — from k2-fsa/sherpa-onnx. These are cached in the plugin folder and reused offline. No code that becomes part of the plugin is ever downloaded.
  • Temporary files. During rendering, the plugin writes temporary audio/frames to your system temp folder and deletes them when finished.

No analytics, no telemetry, no ads. Your generated audio and video never leave your machine.

Requirements

  • Desktop Obsidian (desktop-only — uses Node/Electron APIs; isDesktopOnly).
  • Apple Silicon Mac for the local Kokoro voice (other platforms fall back to the system voice).
  • ffmpeg — auto-downloaded on first use on macOS, or set its path in settings / brew install ffmpeg.
  • An OpenAI or DeepSeek API key (for the script step).
  • (optional) a free Pexels API key (for auto images) — get one at pexels.com/api.

Install

From the community store (once published): Settings → Community plugins → Browse → search ZoyClip.

From source / BRAT:

git clone https://github.com/zoyluoblue/obsidian-note-to-video
cd obsidian-note-to-video
npm install
npm run build        # outputs main.js

Copy main.js, manifest.json, styles.css into <your-vault>/.obsidian/plugins/zoyclip/, enable the plugin in Obsidian, add your API key in settings, then run the command "Turn this note into a vertical short video" on any note.

Settings

  • Language — Auto (follow Obsidian) / English / 中文 — switches UI and output.
  • LLM — provider (OpenAI / DeepSeek), Base URL, model, API key.
  • Voice — Kokoro voice id (audition by changing and re-rendering).
  • Captions — TikTok (big words, pop-in) or full-sentence style.
  • Images — Pexels API key (auto) — or an images folder / ![[note embeds]] (manual).
  • Music — a folder of tracks (random pick + ducking) + volume.
  • Cover — generate a cover image on/off.
  • Preview — edit the script + image keywords before rendering.
  • ffmpeg path — optional (auto-detected / auto-downloaded otherwise).

How it stays local & free

  • Voiceover runs locally via Kokoro + sherpa-onnx — no cloud TTS bills.
  • Rendering uses native ffmpeg plus an in-app Canvas frame renderer that reuses Obsidian's own Chromium — no second browser, no native-binary bundling.
  • Only the script rewrite uses a cloud LLM (your own key); your audio and video never leave your machine.

License

MIT — see LICENSE. Third-party components keep their own licenses: Kokoro (Apache-2.0), sherpa-onnx (Apache-2.0), ffmpeg (downloaded static build, LGPL/GPL per build), and Pexels images under the Pexels License.

For plugin developers

Search results and similarity scores are powered by semantic analysis of your plugin's README. If your plugin isn't appearing for searches you'd expect, try updating your README to clearly describe your plugin's purpose, features, and use cases.