ChenShuo2004/cs-board

What it solves

CS Board is a local AI video production workbench that transforms text scripts into talking videos. It eliminates the need for manual animation and editing by automating the pipeline of voice cloning, content decomposition, illustration generation, and audio-visual synthesis into a final MP4 file.

How it works

The system uses a multi-stage pipeline to generate videos based on user inputs (reference audio, Chinese text, and visual templates):

1. Audio & Content: It integrates with IndexTTS for voice cloning and uses LLMs (like GPT-5) to decompose text into scenes and storyboards.
2. Visuals: It generates illustrations using image models (like GPT-Image-2) based on selected visual templates or custom style/character references.
3. Animation: It creates whiteboard-style hand-drawn animations or dynamic infographics where elements appear in sync with the narration timing.
4. Synthesis: It combines subtitles, audio, and visuals using a rendering engine (Remotion) to export the final video.

Who it’s for

• Content Creators: For those making knowledge-sharing videos, story-telling clips, or course promotions.
• Brand Managers: For creating consistent IP-based videos using custom character and style references.
• Business Analysts: For producing dynamic infographics for commercial analysis or point-of-view expressions.

Highlights

• Local-First Privacy: Assets, API keys, and task history are stored locally on the user's machine.
• 12 Visual Templates: Includes a variety of styles from minimal whiteboard and business doodle to cyber-neon and 3D clay.
• Dynamic Infographics: Synchronizes visual elements with the actual speaking time of the narration to prevent visual "lead-in" errors.
• Voice & Style Libraries: Dedicated management pages for saving and reusing reference voices and visual styles.
• Object-Level Drawing Rhythm: Optimizes the hand-drawn animation process by grouping strokes by object and adjusting timing based on complexity.
• Integrated Image Generator: A standalone tool within the workbench to generate and download PNGs using the configured image models.