Fogsight: An Agent That Generates Animations Through Conversation 199

Fogsight is an open-source animation engine based on LLM that can quickly turn abstract concepts into educational or science-popularization animations with bilingual narration and cinematic visuals, and also allows adjusting details through conversation. It is developed with Python 3.9+, supports local deployment, can be adapted to enterprise privatization needs, and is suitable for education, science popularization, product demonstrations, and other scenarios.

Core Features

1. Input a topic and get an animation directly: For example, by entering "Bubble Sort" or "Law of Entropy Increase," Fogsight will automatically generate a complete teaching animation with both Chinese and English voiceover, saving the trouble of manual production.

2. Full-process automatic orchestration: Scripts, visuals, transition effects, and other steps are all completed automatically by AI, without requiring users to understand animation production.

3. Adjust details through conversation: Through the Language User Interface (LUI), users can say things like "add some colors" or "simplify the narration," and AI will regenerate animations that meet the requirements.

Installation and Usage

Installation Requirements: Python 3.9 or above, common browsers (Chrome/Firefox/Edge), and a large language model API key (Google Gemini 2.5 recommended). Follow the steps to clone the code, install dependencies, configure the key, and start the service to run locally.

Usage Steps:

1. Enter the topic for which you want to generate an animation in the interface (such as "Euler's Theorem");

2. The system automatically generates a first draft with narration, camera scripts, and visuals;

3. Request modifications through conversation (such as "add slow motion" or "adjust color scheme"), and AI adjusts in real time;

4. After previewing and being satisfied, export the video or embed link for use in courseware, social platforms, and other scenarios.

Common Application Scenarios

• Teaching: Teachers enter concepts such as "quantum entanglement" to quickly generate animations with accessible narration, helping students understand complex knowledge;

• Science Popularization: Enter topics such as "black holes" to generate science-popularization videos with visuals, suitable for social media dissemination;

• Technical Training: Enterprises use it to create animations on "blockchain principles" and other topics, visualizing difficult-to-understand technologies;

• Personal Creation: Enter abstract words such as "freedom" or "time" to generate artistic animations for short videos or art projects.

Technical Features

• Full-process automation: Covers 12 steps including script generation and visual design, without the need for manual step-by-step operation;

• Multimodal input: Supports uploading hand-drawn sketches as character references, or using voice descriptions to adjust style;

• Intelligent optimization: Uses algorithms to check animation logic in real time, and automatically adds transitional scenes when broken plots are detected;

• Diverse styles: Built-in with over 200 styles (such as 2D ink painting, 3D cyberpunk), customizable with keywords.

Advantages and Limitations

• Advantages: Much faster than traditional animation production (hours vs. weeks), low cost, easy for non-professionals to get started, and allows adjusting details while rendering.

• Limitations: Creativity may be limited by training data, handling of complex actions (such as multi-character interactions) in terms of body details is not precise enough, and generating high-quality animations requires hardware with strong performance.

Open Source Project:https://github.com/fogsightai/fogsight