What is html-video
html-video is an HTML version of Jianying open-sourced by the Open Design team (nexu.io), built on the hyperframes framework. The tool lets Agents generate world-class product promos, knowledge explainers, and other videos by writing HTML. It features 20+ top-tier templates, supports paginated editing, multi-size export, and Minimax AI voiceover and music, and can run locally without an API Key.
Main features of html-video
- Template-based video generation: Built-in 20+ top-tier HTML video style templates covering product promos, knowledge explainers, data visualization, and other scenarios; Agents can directly call them to generate dynamic MP4s.
- Paginated editing and preview: The Studio editor supports paginated preview, paginated editing, and frame-level text modification, with no need to repeatedly export to check results, greatly improving modification efficiency.
- Multi-size adaptation: One-click switching between mainstream video specifications such as horizontal 16:9, vertical 9:16, square 1:1, and Xiaohongshu 4:5, adapting to distribution needs across all platforms.
- Seamless Agent integration: Automatically recognizes 6 local Code Agent CLIs (Codex, Claude Code, Hermes, Cursor Agent, etc.), with one-click switching in the top bar, ready to use with zero configuration.
- AI voiceover and music: Already integrated with Minimax, supporting automatic generation of narration voiceovers and background music based on video content, with low Token consumption and controllable cost.
How to use html-video
- Clone the repository: Pull the html-video open-source project from GitHub to your local machine.
- Install dependencies: Run the installation command and configure the local environment, with no need to apply for any API Key.
- Launch Studio: Open the local visual editor, browse the 20+ built-in templates, and choose a suitable style.
- Input content: Hand over content such as website links, files, or WeChat official account articles to the Agent, which automatically generates an HTML video script.
- Paginated refinement: Preview and edit text and visuals page by page in Studio, adjusting details in real time.
- Export MP4: After confirming everything is correct, export a high-definition video with one click, or directly generate voiceover and music and then download.
Core advantages of html-video
- Local-first, zero API cost: Runs completely locally, does not rely on cloud API Keys, ensures data privacy, and has extremely low long-term usage cost.
- Agent-native design: Built specifically for Code Agent workflows, supports one-click switching among mainstream CLI tools, and lets AI automatically complete the full chain from content to video.
- HTML-driven, extremely low cost: Uses HTML/CSS to replace traditional video engineering; 30,000 lines of code can achieve a Jianying-level experience, with technical barriers and maintenance costs far lower than professional video software.
- Real-time iteration, WYSIWYG: The paginated editing mode supports frame-by-frame modification and real-time preview, avoiding the inefficient "modify-render-check" loop of traditional video tools.
Project address of html-video
- GitHub repository：https://github.com/nexu-io/html-video
Comparison of html-video with similar competitors
| Dimension | html-video | FireRed-OpenStoryline |
|---|---|---|
| Core positioning | HTML version of Jianying, Agent-native video generation | Conversational AI video editing Agent with director-like thinking |
| Technical architecture | hyperframes + HTML/CSS rendering engine | MoviePy + FFmpeg + LangChain + MCP protocol |
| Interaction mode | Studio visual editor + Agent CLI code-driven | Natural language conversation + real-time Web interface interaction |
| Material sources | User input (links, files, articles) + 20+ built-in templates | Automatic online search (Pexels) + user-uploaded materials |
| Editing granularity | Page preview + frame-level text editing | Conversational fine-tuning (trim, replace, reorganize, color grade) |
| AI capabilities | Template-based generation, Minimax voiceover and music | Intelligent script generation, ASR rough cut, AI transitions, style imitation |
| Agent integration | Native support for 6 CLIs (Codex/Claude Code/Cursor, etc.) | OpenClaw / Claude Code / Codex Skills integration |
| Local/Cloud | Local-first, zero API Key for core features | Local deployment (Docker/Conda), requires LLM API Key configuration |
| Hardware requirements | Runs on an ordinary laptop | No GPU required, core logic runs on CPU |
| Templates/Skills | 20+ built-in HTML video templates | Editing Skill accumulation (reuse of rhythm, color tone, transition styles) |
| Output formats | MP4 in multiple sizes (16:9 / 9:16 / 1:1 / 4:5) | Custom resolution video |
| Open-source license | Apache 2.0 | Apache 2.0 |
| Target users | Developers, Agent workflows, product promotion/knowledge explanation | Content creators, Vlog bloggers, talking-head video producers |
Application scenarios of html-video
- Mass production of content by knowledge bloggers: Automatically convert WeChat official account articles and technical documents into knowledge explanation videos with voiceover, adapting to multi-platform specifications such as Xiaohongshu 4:5 and Bilibili 16:9.
- Product promotion videos: Input a product's official website link, and the Agent automatically extracts selling points and generates a dynamic product showcase video, suitable for startups to quickly produce marketing materials.
- Data visualization reports: Transform CSV/JSON data into dynamic chart videos for presentation, social media distribution, and to avoid the dullness of static charts.
- E-commerce content creation: Batch-generate short videos such as product introductions and usage tutorials, paired with Minimax AI voiceover, to achieve low-cost, high-efficiency content matrix operations.