Fogsight is a free and open source AI animation generation engine. It is driven by LLM. By inputting abstract concepts or keywords, it can automatically generate HTML animations containing narration, visual elements and dynamic effects. It supports multi-round dialogue optimization and multiple LLM backends.

Fogsight

Core parameters and statistics

Project	Specifications
Product name	Fogsight
Category	AI video generation
Delivery form	Web / Self-hosted
Support Platform	Web, Desktop
Supported languages	zh-CN, en-US
Target Users	Educators, content creators, developers
User scale	Initial growth stage of the community
Pricing model	Freemium (free quota + subscription)

Platform coverage and user scale data are based on the official real-time page and third-party statistics.

User and market recognition

Fogsight is incubated by the WaytoAGI open source community, and its core contributors come from universities such as the Chinese University of Hong Kong (Shenzhen), Nanyang Technological University, and the University of Southern California. The project is open sourced on GitHub, and community activity continues to grow, attracting the attention of educators, popular science creators, and technology developers. After the SaaS version (fogsight.ai) is launched, the Free plan opens a daily free quota to individual users and has accumulated an initial user base. Pro and Premium paid plans are aimed at high-frequency creators and team users. The value points most recognized by users are the zero-threshold experience of "entering the topic and getting the complete animation" and the flexible control capabilities brought by the LUI's multi-round dialogue optimization. The product belongs to the "concept visualization" category in the AI ​​video generation track, forming misaligned competition with tools such as Pika and Runway that focus on realism.

Cost advantage

Cost Dimension	Description
Free version	Register at fogsight.ai to get 50 Credits per day (approximately 5 times generated)
Subscription version	Pro $9.9/month (1500 Credits), Premium $29.9/month (6000 Credits)
Enterprise Edition	Unpublished, need to contact the team to confirm commercial authorization

Compared with traditional animation outsourcing (hundreds to thousands of yuan per unit) or professional animation software (such as Adobe After Effects $20.99/month + learning cost), Fogsight compresses the cost of a single animation to almost zero (within the free quota). Self-hosted users only need to bear the cost of LLM API calls, which is about $0.01-$0.05 each time using Gemini 2.5 Pro as an example.

Main functions

• One-click generation from concept to animation: Enter a topic or concept (such as "bubble sort"), and automatically call LLM to generate a complete animation script and HTML animation, including bilingual narration, movie-level visual texture and dynamic transition, merging the traditional five-step process into one step.
• LLM-driven intelligent orchestration engine: The orchestration layer converts user input into animation description, independently determines visual elements, layout, animation timing and narration rhythm. The narrative structure and visual style generated each time are dynamically adapted instead of applying templates.
• Language User Interface (LUI) Multiple Rounds of Tuning: Iteratively tune animations through natural language dialogue, such as "make the narration more colloquial" or "slow down the tempo of the second paragraph" without understanding the concept of timelines or keyframes.
• Multi-model backend compatibility: Supports multiple LLM backends such as OpenAI GPT, Google Gemini, OpenRouter, etc. When self-hosting, you only need to modify the API_KEY and MODEL fields in the configuration file to switch.

Model and version evolution

Version	Date	Key Changes
v1.0.0 (public version)	~2025-12	LLM intelligent orchestration engine, LUI multi-round tuning, multi-model compatibility, Docker deployment
v0.5 (early preview version)	~2025-09	Basic concept to animation conversion link, only supports OpenAI GPT

The version record shall be subject to the official release notes.

Technical advantages

• LLM orchestration + code generation two-layer architecture: The orchestration layer uses LLM's semantic understanding capabilities to handle complex concepts, and the rendering layer focuses on generating high-performance HTML/SVG/JS animations. The two are decoupled and can be upgraded independently.
• Dynamic adaptation instead of template application: The narrative structure, visual composition and motion rhythm of each generated animation are the result of dynamic arrangement by LLM based on the input concept. The same theme has different generated effects each time.
• Lightweight self-hosting and SaaS dual-mode deployment: Supports Docker one-click deployment, self-hosting only requires LLM API Key; fogsight.ai provides hosting services without local configuration.
• Cross-model compatibility: With OpenRouter compatible with multiple LLM backends, users can choose models based on cost and performance preferences.

How to use

Entrance	How to use
Web client (SaaS)	Visit fogsight.ai → Enter Workspace → Enter the topic → Generate → LUI Tuning → Export
Self-hosted (open source)	git clone → pip install → configure credentials.json → python start_fogsight.py
Docker	docker-compose up -d → visit localhost:8000

Typical usage process: Register account → Enter concept → AI automatic arrangement → View animation → LUI tuning → Export HTML/MP4.

Product Pricing

Package	Price	Contents
Free	$0/month	50 Credits/day (~5 builds), base model, community support
Pro	$9.9/month	1500 monthly Credits + 50-day refresh, premium model, priority support
Premium	$29.9/month	6000 monthly Credits + 50-day refresh, unlimited devices, early access

Pricing. Self-hosted users have zero platform fees with the open source version and only bear the cost of the LLM API.

Application scenarios

• Education and Teaching: Teachers input knowledge points to automatically generate teaching animations, improving students' understanding efficiency by 30%-50%. It is suitable for science and engineering courses from K12 to higher education.
• Popular science content creation: Popular science bloggers input scientific concepts to quickly generate explanatory animations. The automatically generated bilingual narration is suitable for bilingual content creation scenarios in Chinese and English.
• Corporate Training and Knowledge Sharing: Input management concepts to generate standardized training animations. When knowledge is updated, it only needs to be regenerated rather than rerecorded.

Applicable people

• Individual users: Educators, popular science creators, and students can experience it at zero cost through free quota or self-hosting.
• SME Team: Enterprise training and knowledge management teams need a standardized and reusable training material library.
• Large Enterprises: Self-hosted deployment can be run on the enterprise intranet to avoid the risk of data transmission, but the CC BY-NC-ND 4.0 commercial authorization path needs to be confirmed.
• Not suitable for boundaries: Scenes that require extremely high image quality or hand-painted style (limited visual style unity); customized brand assets are required (customized brand color palettes are not supported).

Comparison of competing products

Comparative Dimensions	Fogsight	Pika	Runway
Core differences	Animation of knowledge concepts	Realistic video generation	Professional video editing
Price	Free + Starting at $9.9/month	Starting at $10/month	Starting at $15/month
Covered scenes	Educational science popularization, concept visualization	Social media, creative videos	Film and television production, advertising
User evaluation	Zero threshold, education-friendly	High visual quality	Rich functions
Technical threshold	Low (no animation skills required)	Low	Medium

Summary and Outlook

Fogsight takes "concepts as images" as its core philosophy, combines LLM semantic understanding with animation code generation, and has a unique positioning in the fields of educational science popularization and knowledge visualization. Current limitations include limited animation style diversity, reliance on screen recording for MP4 export, and arrangement quality limited by LLM capabilities. The open source license (CC BY-NC-ND 4.0) prohibits unauthorized commercial use, and enterprise users need to obtain separate authorization.

Acquisition/Adoption Risk Assessment: Individual users and educators can start with the Free plan or self-hosted version with confidence. Before adoption, enterprises need to confirm whether the business authorization path is clear and whether sensitive data is acceptable for transmission through the LLM API. It is recommended to use the SaaS Free plan to verify the match between the product and the scenario before deciding on the deployment plan.

Parameter item	Description
Product form	Open source web application (self-hosted) + cloud SaaS (fogsight.ai)
Core capabilities	Automatic generation of LLM from concept to animation, intelligent orchestration, and multiple rounds of language user interface (LUI) tuning
Supported LLM backends	OpenAI GPT, Google Gemini, OpenRouter (compatible with Anthropic Claude, etc.)
Output format	HTML (standalone), MP4 video (export via screen recording)
Runtime context	Python 3.10+, supports Docker deployment
License	CC BY-NC-ND 4.0 (Open Source for Non-Commercial Use)
Pricing model	Free quota + subscription (Pro $9.9/month Premium $29.9/month)

Each generation costs 10 Credits. The free version refreshes 50 Credits daily (about 5 times generated), the Pro version includes an additional 1500 monthly Credits, and the Premium version includes 6000 monthly Credits. Compared with traditional animation production software (such as After Effects, Blender), Fogsight compresses the production time of a single animation from days to seconds. The price is that the customization space of animation style and details is limited by the orchestration capabilities of LLM.

User and market recognition

Fogsight is incubated by the WaytoAGI open source community, and its core contributors come from universities such as the Chinese University of Hong Kong (Shenzhen), Nanyang Technological University, and the University of Southern California. The project is open sourced on GitHub, and community activity continues to grow. The direction of AI-driven animation generation has attracted the attention of educators, popular science creators, and technology developers.

After the SaaS version (fogsight.ai) is launched, the Free plan opens a daily free quota to individual users and has accumulated an initial user base. Pro and Premium paid plans are aimed at high-frequency creators and team users. The product has not yet disclosed specific user numbers or enterprise cooperation cases, but its "free open source + cloud service value-added" model has lowered the adoption threshold in the education and science field. Judging from community feedback, the value points most recognized by users are the zero-threshold experience of "entering the topic and getting the complete animation" and the flexible control capabilities brought by the LUI's multi-round dialogue optimization.

This product belongs to the "conceptual visualization" category in the AI ​​video generation segment, and forms misaligned competition with tools such as Pika and Runway that focus on realistic/stylized video generation. Fogsight does not pursue pixel-level visual realism, but focuses on converting abstract knowledge into narrative animation, which is more suitable for education, training and science popularization scenarios. It is recommended to refer to the GitHub Issues discussion and third-party evaluation platform for user reputation and industry adoption.

Cost advantage

Fogsight’s cost structure is divided into three tiers:

• C-side users: Free plan provides 50 credits per day (approximately 5 times generated), Pro plan $9.9/month (approximately ¥69) provides 1500 monthly credits + 50 daily refresh credits, Premium plan $29.9/month (approximately ¥199) provides 6000 monthly credits + 50 daily refresh credits. Extra Credit Package $1.90/300 Credits. Compared with traditional animation outsourcing (single animation is hundreds to thousands of yuan) or professional animation software subscription (such as Adobe After Effects $20.99/month + learning cost), Fogsight compresses the cost of a single animation to almost zero (within the free limit) or several yuan.
• Developers/Self-hosted users: The project is completely open source, with zero platform fees for self-hosting, and only needs to bear the cost of LLM API calls. Taking Google Gemini 2.5 Pro as an example, the API cost for each animation generation is about $0.01-$0.05, which is much lower than a commercial subscription.
• Enterprise/Team Users: There is currently no public enterprise privatization plan and pricing. If you need commercial use (CC BY-NC-ND 4.0 prohibits unauthorized commercial use), you need to contact the team to obtain a commercial license.

Cost Dimension	Description
Free trial	Register at fogsight.ai to get 50 Credits per day (approximately 5 times generated)
Personal subscription	Pro $9.9/month (1500 Credits), Premium $29.9/month (6000 Credits)
Self-hosted	Open source and free + LLM API pay-as-you-go (Gemini 2.5 Pro is about $0.01-$0.05/time)
Enterprise/Commercial Authorization	Unpublished, need to contact the team for confirmation

The core of the cost advantage lies in compressing the traditional multi-role workflow of "copywriting scripting + visual design + animation production + dubbing and narration" into a single AI orchestration, eliminating the cost of professional software licensing, outsourcing fees and cross-role communication. This cost structure has significant appeal for educational institutions and content teams that need to produce explanatory animations at scale.

Main functions

• One-click generation from concept to animation: The user inputs a topic or concept (such as "bubble sorting", "entropy increase law"), and Fogsight automatically calls LLM to generate a complete animation script (including narration, visual storyboards, and dynamic effects), without the need to manually design any animation details. The result is a ready-to-run HTML file with bilingual narration (Chinese + English), cinematic visual textures, and smooth dynamic transitions. This process combines the five-step process of "creative conception → script writing → storyboard design → animation production → dubbing synthesis" in traditional animation production into one step.
• LLM-driven intelligent orchestration engine: Fogsight’s core differentiating capability is its LLM-driven orchestration layer. This orchestration layer is not only responsible for converting user input into animation scripts, but also independently determines the selection, layout, animation timing, and narration speed and rhythm of visual elements. The orchestration engine outputs a structured animation description, which is then converted into actual HTML/SVG/JS code by the rendering engine. This means that each generated animation is dynamically adapted in narrative structure and visual style, rather than applying a fixed template.
• Language User Interface (LUI) multiple rounds of tuning: The generated animation is not finalized in one go. Users can conduct multiple rounds of conversations with Fogsight through natural language, and accurately tune the narration wording, visual style, rhythm and speed of the animation. For example, the user can say "make the narration more colloquial" or "slow down the animation pace of the second paragraph", and the system will make incremental adjustments based on the original. This LUI interaction mode lowers the technical threshold for animation editing - users do not need to understand timeline, keyframe or layer concepts, and only need to describe modification intentions in everyday language.
• Multi-model backend compatibility: Fogsight supports access to multiple LLM backends, including OpenAI GPT, Google Gemini (Gemini 2.5 Pro recommended) and OpenRouter (can access Anthropic Claude, Meta Llama

wait). Users have the flexibility to choose models based on cost, quality and preference. In self-hosted mode, simply modify the API_KEY, BASE_URL, and MODEL fields in credentials.json to switch backends.

• Various Export and Sharing: Animations can be exported as independent HTML files (open and run in any browser), and MP4 video format export is supported (through the browser screen recording function). The exported HTML file is small in size and has zero dependencies, making it suitable for embedding into websites, courseware, or sharing with others. The MP4 video export function is still under development.

Model and version evolution

Fogsight’s version evolution is highly coupled with the iteration of the underlying LLM capabilities—each version update often corresponds to an upgrade of the orchestration engine or the integration of a new model backend.

Version	Release Date	Main Features
v0.5 (early preview version)	~2025-09	Conversion link from basic concepts to animation, supporting OpenAI GPT
v1.0.0 (public version)	~2025-12	LLM intelligent orchestration engine, language user interface (LUI) multiple rounds of tuning, multi-model compatibility (GPT / Gemini / OpenRouter), HTML export Docker deployment support

The project continues to be actively developed on GitHub, and the latest merged PR is #22 (around 2026-03). Function iteration directions include: improving the stability of animation code generation, expanding the editing capabilities of LUI, and optimizing the MP4 video export process. There is currently no official semantic version release tag (GitHub Releases is empty). The above version number comes from the functional milestone description in the project README. The underlying model update (such as the access to Gemini 2.5 Pro) is synchronized with the adaptation of the orchestration engine. It is recommended to pay attention to the GitHub submission log for the latest updates.

Technical advantages

• Dual-layer architecture of LLM orchestration + code generation: Fogsight does not simply use LLM output as an overlay for the video, but builds a two-layer pipeline - the first layer (arrangement layer) is used by LLM to analyze user input and generate structured animation descriptions (including narration text, visual element definitions, animation timing, scene switching logic); the second layer (rendering layer) converts structured descriptions into executable HTML/SVG/JS code. The advantage of this architecture is that the orchestration layer can make full use of LLM's semantic understanding capabilities to process complex concepts, while the rendering layer focuses on generating high-performance web page animations. The decoupling of the two allows either layer to be upgraded independently.
• Dynamic adaptation rather than template application: Unlike most "text-to-video" tools that rely on preset templates to reduce the complexity of generation, the narrative structure, visual composition, and motion rhythm of each animation generated by Fogsight is the result of dynamic arrangement by LLM based on the input concept. This means that the same topic may generate different results each time, and the system will select the most appropriate representation based on the characteristics of the concept (for example, mathematical theorems are suitable for graphic display, and historical events are suitable for timeline narratives).
• LUI interaction lowers the editing threshold: Language User Interface (LUI) is Fogsight's core innovation in "editing after generation". Traditional animation tools require a separate timeline editor to adjust animation details, while Fogsight allows users to complete iterative optimization through natural language dialogue. The principle is to send the user's modification instructions and the original animation description to LLM, and LLM generates the updated animation description and then re-renders it, forming a "generation → feedback → regeneration" structure.
• Lightweight self-hosting and SaaS dual-mode deployment: The project supports one-click deployment of Docker (docker-compose up -d), and it can run with Python 3.10+. Self-hosted users only need an LLM API Key to get full animation generation capabilities

force. For users who don’t want to deal with infrastructure, fogsight.ai offers a managed SaaS service that doesn’t require any on-premises configuration.

• Cross-model compatibility: By accessing OpenRouter, Fogsight is compatible with a variety of LLM backends, allowing users to select models based on cost and performance preferences. Google Gemini 2.5 Pro is recommended for a balance between cost and quality.

How to use

Fogsight provides two usage paths: Cloud SaaS (zero deployment) and Open source self-hosting (requires local environment).

Cloud SaaS (recommended for newbies)

1. Visit https://fogsight.ai/ and click "Enter Workspace" to enter the workspace.
2. Enter the topic or concept you want to animate in the input box (such as "Bubble Sort", "Law of Entropy Increase").
3. Click Generate and wait for AI to automatically complete animation arrangement and rendering (about 10-30 seconds).
4. View the generated HTML animation and perform multiple rounds of optimization through the LUI dialog window.
5. Click "Save as HTML" to download a standalone HTML file, or "Export as Video" to export to MP4.

Open source self-hosting

Contextual requirements: Python 3.10+, modern browser LLM API Key (Google Gemini 2.5 Pro recommended)

# 1. Clone the repository
git clone https://github.com/fogsightai/fogsight.git
cd fogsight

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure API key
cp demo-credentials.json credentials.json
# Edit credentials.json and fill in API_KEY and BASE_URL
# Gemini configuration example:
# {
# "API_KEY": "<YOUR_GEMINI_API_KEY>",
# "BASE_URL": "",
# "MODEL": "gemini-2.5-pro"
# }

# 4. One-click start
python start_fogsight.py
# The browser automatically opens http://127.0.0.1:8000

# 5. Start creating

Docker way:

git clone https://github.com/fogsightai/fogsight.git
cd fogsight
cp demo-credentials.json credentials.json
# Edit credentials.json to configure API keys
docker-compose up -d
# Visit http://localhost:8000

Product Pricing

Fogsight uses a Credits system for billing, and each animation generation consumes 10 Credits. Supports Stripe (credit card) and WeChat payment.

Package	Price	Credits	Number of Devices	Features
Free	$0/month	50 Credits/day (about 5 times/day)	1 unit	Basic model, community support
Pro	$9.9/month (≈¥69)	1500 monthly + 50-day refresh	2 units	Advanced model, priority support
Premium	$29.9/month (≈¥199)	6000 monthly + 50-day refresh	Unlimited	Advanced models, early access to new features
Credit Top-up	$1.90/pack (≈¥13)	300 Credits/pack, valid for 30 days	—	Pro/Premium subscription required, stackable

Credit consumption sequence: Daily refresh Credits → Pro subscription Credits → Premium subscription Credits → Recharge package Credits. All subscriptions can be canceled at any time and will expire after the current period. The upgrade takes effect immediately, and the downgrade takes effect in the next cycle.

Self-hosted users do not need to pay platform fees to use the open source version, only the cost of LLM API calls. Commercial use requires contacting the team to obtain authorization (the open source license is CC BY-NC-ND 4.0, unauthorized commercial use and modification and derivatives are prohibited).

Application scenarios

• Education and Teaching: Teachers input knowledge points such as "Pythagorean Theorem" or "Cell Mitosis" and automatically generate teaching animations to assist classroom explanations. Compared with static PPT, animated knowledge display can improve students' understanding efficiency by 30%-50% (data from the field of education research), and teachers do not need to have animation production skills. Science, engineering and humanities courses suitable for K12 to higher education levels.
• Popular science content creation: Popular science bloggers input scientific concepts such as "quantum entanglement" or "CRISPR gene editing" and quickly generate explanatory animations for short videos or social media dissemination. Fogsight's automatically generated bilingual narration makes it naturally suitable for bilingual content creation scenarios in Chinese and English.
• Corporate Training and Knowledge Sharing: Corporate trainers input management concepts such as "SWOT Analysis" or "AARRR Funnel Model" to generate standardized training animations for internal knowledge base or sharing demonstrations. Animated formats are easier to maintain and update than recorded live training videos—when knowledge is updated, it simply needs to be regenerated rather than rerecorded.
• Academic Demonstration and Thesis Defense: Graduate students and scholars input core research concepts (such as "Transformer attention mechanism"), and generate concept animations to embed in presentations or supplementary materials for thesis. Animated concept explanations in academic reports can significantly improve reviewers' understanding of complex theories.
• Product Concept Proof (MVP): Entrepreneurs or product managers input the core concept of the product and generate concept animations for early prototype display or investor communication, avoiding high-cost video production outsourcing during the concept proof stage.

Applicable people

• Educators (Teachers/Professors/Trainers): Need to visualize abstract knowledge points to improve the teaching effect of classroom or online courses. Fogsight's zero-barrier operation makes it ideal for non-technical educators. Not suitable for boundaries: When extremely high-quality or traditional hand-drawn style animation is required (such as art courses), the visual style uniformity of the current LLM arrangement is limited.
• Science creators and self-media people: To produce short knowledge videos for the public, it is necessary to produce explanatory animation content in batches and efficiently. The Free plan generates 5 times a day to basically meet daily updates. Not suitable for boundaries: For short video types that require real-life appearances or a mixture of real-life footage, Fogsight only generates pure animations.
• Open source enthusiasts and technology developers: Developers who are concerned about the combination of AI Agent and code generation can conduct secondary development based on Fogsight's open source code and integrate its animation orchestration capabilities into their own projects or workflows. Prerequisites: Experience in Python and LLM API is required.
• Students and self-learners: Transform difficult concepts in learning into animations to assist understanding, especially suitable for visual learners. The self-hosted version costs zero (only requires an API Key), making it suitable for students on a budget.
• Corporate Training and Knowledge Management Team: A standardized and reusable training material library is needed. Through self-hosted deployment, it can be run on the enterprise intranet to avoid the risk of data external transmission. Not suitable for boundaries: Scenarios that require highly customized visual brand assets (strict specifications for enterprise VI). The current version does not support custom brand color palettes and font systems.

Summary and Outlook

Fogsight takes "concepts as images" as its core philosophy and combines LLM's semantic understanding capabilities with animation code generation technology to provide the shortest path from abstract concepts to concrete animation. Its open source attributes, multi-model compatibility and LUI interaction mode give it a unique product positioning in the fields of education science popularization and knowledge visualization - not directly competing with general video generation tools (such as Pika, Runway), but opening up the vertical scenario of "knowledge animation".

Current Limitations and Uncertainties:

• Limited diversity of animation styles: The animations generated by the current version are mainly infographic-style explanations, with a high degree of uniformity in visual style and a lack of realistic/artistic style options.
• MP4 video export relies on browser screen recording, and the experience is not smooth enough. The official is developing native video export capabilities.
• Choreography quality is subject to LLM capabilities: when using weaker models (such as the free version default model), the narrative logic and visual presentation of the animation may be biased.
• Open source license restrictions: CC BY-NC-ND 4.0 prohibits commercial use and modification and derivatives. Enterprise users need to obtain separate authorization for commercial use.
• There is currently no official mobile application, and all interactions rely on web browsers.

Acquisition/Adoption Risk Assessment: Individual users and educators can start with the Free plan or the self-hosted version at a minimal cost. Enterprises need to confirm before considering adoption: (1) Whether the commercial authorization path under the CC BY-NC-ND 4.0 license is clear; (2) When the generated animation content involves sensitive enterprise information, whether the data is accepted to be transmitted through the LLM API; if data localization is required, the self-hosted model can avoid this risk, but it needs to manage the LLM API key and infrastructure by itself. It is recommended to first use the SaaS Free plan to verify the match between the product and the scenario, and then decide whether to adopt a self-hosted or commercially authorized deployment solution.

Related tools: runway, pika

Version Info

• public version ：Supports automatic generation from concept text to complete animation, including LLM orchestration engine, language user interface (LUI) multi-round optimization, multi-model compatibility (GPT / Gemini / OpenRouter), and HTML export. There is no official precise date yet. （~2025-12）
• early preview ：The first runnable version supports the conversion link from basic concepts to animations and is only compatible with OpenAI GPT. There is no official precise date yet. （~2025-09）