Paper2Any
Turn paper/text/topic into editable research figures, technical route diagrams, and presentation slides.
Last checked:
What is Paper2Any?
Paper2Any is an open-source multimodal AI platform developed by Peking University's DCAI research group that transforms academic papers, text, images, and topics into editable scientific figures, technical diagrams, and presentation slides. It automates the entire visualization workflow from reading materials and understanding logic to generating images, cutting elements, and outputting fully editable PPTX and SVG files.
The tool offers 13+ core capabilities including Paper2Figure (model architecture diagrams, technical roadmaps, experimental plots with editable PPTX output), Paper2PPT (paper/text/topic to slide decks with long-doc support), PDF2PPT (layout-preserving conversion), Image2PPT (image to slides), Paper2Diagram/Image2Drawio (editable draw.io diagrams), PPTPolish (AI-based layout optimization and style transfer), Paper2Video (video script generation), Paper2Poster (academic poster creation), Paper2Rebuttal (rebuttal drafting), Paper2Citation (citation explorer), Paper2Technical (technical reports), Knowledge Base workflows, and an Image Model Playground for batch image generation.
Paper2Any is designed for researchers, academics, students, and anyone who needs to create professional presentations and scientific figures from research papers. It supports universal input formats (PDF, images, text, topics), produces editable outputs with independent elements (blocks, arrows, text), and allows customizable visual styles while preserving original assets.
Paper2Any pricing
Pricing model: Freemium
Paper2Any is free and open-source under Apache 2.0 license. The web demo offers free registration for new users with daily quota allocation through a user points management system. After logging in, users can view historical usage records. The self-hosted version is completely free but requires users to provide their own LLM API keys (e.g., from OpenAI, Anthropic, DeepSeek, or APIYI gateway which supports 400+ models). Core features work without Supabase authentication, but cloud features like points, redeem codes, invites, and history require Supabase configuration. APP_BILLING_MODE=free is the default setting for the free tier.
Paper2Any pros
- Open-source and free to use with Apache 2.0 license
- Generates fully editable PPTX and SVG outputs, not flat images
- Supports universal input: PDFs, images, text, and topics
- One-click generation for scientific figures and slides
- AI-powered layout optimization and style transfer
- Preserves original layout and assets during conversion
- Long document support for 40+ slide presentations
- Intelligent table and figure extraction from papers
- Multiple diagram types: model architecture, technical roadmaps, experimental plots
- Draw.io support for visual diagram creation and chat-based edits
- AI-assisted outline editing with targeted rewrite prompts
- Daily quota allocation with user points management system
- Supports 400+ LLM models including GPT-4o, Claude, Gemini, DeepSeek, Qwen
- Dynamic model switching in web interface without code changes
- History use records viewable after free registration
- Drag-and-drop upload with real-time progress display
- Version history and iterative deck management
- Knowledge base with semantic search for KB-driven PPT generation
- Paper2Rebuttal for structured response guidance in revisions
- ONLYOFFICE online editing support for editable PPTX
Paper2Any cons
- Requires LLM API key configuration (e.g., DeepSeek, GPT) for most features
- PDF2PPT, Image2PPT, and Image2Drawio require SAM3 segmentation service
- Complex deployment requiring Python 3.11/3.12, Conda, and multiple system dependencies
- Needs external system tools: ffmpeg, LibreOffice, Inkscape, poppler-utils, wkhtmltopdf, tectonic
- GPU required for optimal performance with CUDA 12 support
- SAM3 segmentation service requires separate GPU server setup or external endpoint
- Supabase configuration needed for user authentication and cloud features
- First Docker build takes significant time due to system and Python dependencies
- Learning curve for environment variable configuration and model setup
- Web demo URL uses cpolar internal tunnel which may have availability issues
Frequently asked questions about Paper2Any
What input formats does Paper2Any support?
Paper2Any supports universal file formats including PDF files (paper PDFs), images/screenshots, plain text, LaTeX source code, and topics. Users can upload paper PDFs, screenshots of figures, text descriptions, or even just research topics to generate visual outputs.
What output formats does Paper2Any generate?
Paper2Any produces fully editable outputs including PPTX (PowerPoint) files, SVG diagrams, draw.io files, PNG images, and HTML-based editable PPTX. The outputs contain independent editable elements (blocks, arrows, text) rather than flat images, allowing users to modify content after generation.
Is Paper2Any really free to use?
Yes, Paper2Any is open-source under Apache 2.0 license and completely free. The web demo offers free registration with daily quota through a points system. Self-hosted version is free but requires users to bring their own LLM API keys for the AI models.
What LLM models does Paper2Any support?
Paper2Any supports 400+ models including GPT-4o, Claude Sonnet 4.5, Gemini, DeepSeek, Qwen, and many others through APIYI or OpenAI-compatible API format. Models can be dynamically switched in the web interface without code changes via a three-layer model configuration system.
How do I deploy Paper2Any locally?
The recommended method is Docker: clone the repository, configure .env files (simple mode recommended), and run bash deploy/docker-up.sh. Alternative Linux installation uses Conda with Python 3.11, installs dependencies from requirements files, and requires system tools like ffmpeg, Inkscape, LibreOffice. Windows installation is also supported but Linux/WSL is recommended.
What is the difference between Paper2Figure, Paper2PPT, and PDF2PPT?
Paper2Figure generates editable scientific figures (model architecture diagrams, technical roadmaps, experimental plots). Paper2PPT converts papers/text/topics to slide decks with long-doc support and AI enhancement. PDF2PPT performs layout-preserving PDF to PPTX conversion focused on accurate layout retention rather than AI enhancement.
Do I need a GPU to run Paper2Any?
GPU is not required for features that only need LLM APIs (Paper2PPT, Paper2Figure, Knowledge Base). However, PDF2PPT, Image2PPT, and Image2Drawio require SAM3 segmentation service which needs GPU. Docker starts backend + frontend by default without GPU services unless DOCKER_WITH_SAM3=1 is set.
Can I edit generated slides after creation?
Yes, Paper2Any generates fully editable PPTX files with independent elements. The web interface also supports AI-assisted outline editing with targeted rewrite prompts, canvas editing while keeping deck theme locked, and structured outline editing down to section and bullet detail.
What are the system requirements for self-hosting?
Python 3.11/3.12, Conda environment, system tools (ffmpeg, inkscape, libreoffice, poppler-utils, wkhtmltopdf, tectonic LaTeX engine), and optionally NVIDIA GPU with CUDA 12 for advanced features. Ubuntu/Debian is recommended. For Docker deployment, Docker and docker-compose are required.
How does the Knowledge Base feature work?
The Knowledge Base supports document ingestion with embeddings, semantic search capabilities, and KB-driven content generation including PPT, podcast, and mindmap creation. It enables retrieval-augmented workflows where generated content is grounded in uploaded documents and knowledge base context.