Skip to main content

​
Overview

Paper2Any is an open-source multimodal workflow platform for academic papers. It converts paper PDFs, screenshots, or text into model architecture diagrams, technical roadmaps, experimental plots, PPT presentations, and more — all with one click.
Project Info
  • 🔗 Source Code: github.com/OpenDCAI/Paper2Any
  • 📜 License: Open Source
  • 👤 Organization: OpenDCAI
  • ⭐ Community contributed, supports multiple LLMs via APIYI

​
Why Paper2Any

Multiple Output Formats

Convert papers to architecture diagrams, roadmaps, PPTs, rebuttals, and more — one tool for the entire research workflow

Flexible Model Selection

Dynamically switch between GPT-4o, Claude Sonnet, Qwen-VL and more via API parameters — no hardcoding needed

CLI + Web Dual Mode

Both command-line scripts and web interface available to suit different workflows

OpenAI-Compatible API

Native support for OpenAI-compatible API format — just configure APIYI’s Base URL to access 200+ models

​
Core Modules

ModuleDescriptionOutput Formats
Paper2FigureGenerate scientific visualizations from papersArchitecture diagrams, roadmaps (PPTX + SVG), plots
Paper2DiagramCreate flowcharts from papers/text/imagesdraw.io / PNG / SVG
Paper2PPTConvert papers to presentationsPPTX (supports 40+ slides)
Paper2RebuttalGenerate structured rebuttal responsesRebuttal docs with evidence grounding
PDF2PPTLayout-preserving PDF to editable PPTPPTX
Image2PPTTransform images/screenshots into slidesPPTX
PPTPolishAI-driven layout optimizationPPTX
Knowledge BaseFile ingestion, semantic search, KB-driven generationMultiple formats

​
Connect to LLMs via APIYI

Paper2Any supports OpenAI-compatible API format. After configuring APIYI as the LLM endpoint, you can use GPT, Claude, Gemini, DeepSeek, and 200+ other models.

​
Docker Deployment

1

Step 1: Get Your APIYI API Key

  1. Visit APIYI Console to register/login
  2. Go to the Tokens section
  3. Generate a new API key
  4. Copy the key (starts with sk-)
2

Step 2: Clone and Configure Backend

After cloning the repository, edit fastapi_app/.env to set APIYI as the LLM endpoint:
# fastapi_app/.env
DEFAULT_LLM_API_URL=https://api.apiyi.com/v1
BACKEND_API_KEY=sk-your-apiyi-key
Optionally, specify default models for different workflows:
PAPER2PPT_DEFAULT_MODEL=gpt-4o
PDF2PPT_DEFAULT_MODEL=gpt-4o
3

Step 3: Configure Frontend

Edit frontend-workflow/.env to default the web UI to APIYI:
# frontend-workflow/.env
VITE_DEFAULT_LLM_API_URL=https://api.apiyi.com/v1
VITE_LLM_API_URLS=https://api.apiyi.com/v1
4

Step 4: Launch

Start everything with Docker Compose:
docker compose up -d --build
Once started, open the frontend to begin using Paper2Any.

​
CLI Usage

Paper2Any provides standalone CLI scripts with --api-url and --api-key parameters for direct APIYI integration: 
# Paper to PPT
python script/run_paper2ppt_cli.py \
  --input paper.pdf \
  --api-url https://api.apiyi.com/v1 \
  --api-key sk-your-apiyi-key \
  --model gpt-4o

# Paper to Figure
python script/run_paper2figure_cli.py \
  --input paper.pdf \
  --api-url https://api.apiyi.com/v1 \
  --api-key sk-your-apiyi-key \
  --graph-type model_arch
Model Recommendations: For paper-to-PPT conversion, GPT-4o or Claude Sonnet 4.5 are recommended for their strong long-document understanding and structured output capabilities. For diagram generation, vision models like Qwen-VL are also worth trying.

​
Deployment Options

MethodRequirementsBest For
Docker (Recommended)One-click frontend + backend startupQuick start, production
Linux NativePython 3.11+, LaTeX, Inkscape, LibreOfficeDevelopment, customization
WindowsPython 3.12, InkscapeLocal use
GPU-dependent features like PDF2PPT and Image2PPT require a separate SAM3 model server. See the project README for GPU deployment instructions.

​
FAQ

Set DEFAULT_LLM_API_URL to https://api.apiyi.com/v1 and BACKEND_API_KEY to your APIYI key in the environment variables. For CLI mode, use the --api-url and --api-key parameters.
Via APIYI, you can access 200+ models including GPT-4o, Claude Sonnet 4.5, Gemini, DeepSeek, Qwen, and more. Models can be dynamically switched in the web interface without code changes.
Verify that:
  1. Docker and Docker Compose are properly installed
  2. .env files are correctly configured
  3. Required ports are not in use
  4. Check docker compose logs for detailed error messages
  • Ensure your APIYI account has sufficient balance
  • For long papers, use models with larger context windows (e.g., GPT-4o 128K)
  • Check that the paper PDF is searchable text (scanned PDFs may produce poor results)
Visit the APIYI Console, create an account, and generate a new key in the Tokens section. New users receive free trial credits.

APIYI Model List

View the complete list of 200+ models supported by APIYI

Base URL Configuration

Learn how to configure APIYI Base URL in various tools

APIYI Token Management

Manage API keys, check usage and balance

APIYI Pricing

View model pricing and top-up offers