What it does Paper2Any is a Python-based web app that ingests academic papers—PDFs, images, or plain text—and generates editable outputs: model architecture diagrams, experimental plots, technical roadmaps, slide decks, rebuttal drafts, posters, and even video scripts. It exports to PPTX, SVG, and draw.io formats rather than static images, so you can still tweak the results.

The interesting bit The project bundles a surprising breadth of academic workflows under one roof. The draw.io integration is the stand-out: upload a screenshot of a figure and the system reconstructs it as an editable diagram you can refine via chat. There’s also a full frontend canvas for slide editing with outline assistance, gallery review, and version history—more like a lightweight Canva bolted onto an LLM pipeline than a simple script.

Key highlights

• Paper2Figure: Generates model diagrams, technical roadmaps, and experimental plots with editable PPTX/SVG output
• Image2Drawio: Converts paper figures or screenshots into editable draw.io diagrams with chat-based refinement
• Paper2PPT: Long-document PPT generation with built-in table/figure extraction and a frontend canvas editor
• Paper2Rebuttal: Structured rebuttal drafting with claims-to-evidence grounding
• Dynamic model selection: Supports GPT-4o, Qwen-VL, and others via API parameters without hardcoded dependencies
• Knowledge Base: Semantic search and KB-driven generation for PPTs, podcasts, and mindmaps
• Image Model Playground: Batch generation (1/2/4/8/16 images) with compressed previews and zip download

Caveats

• The online demo runs on a personal NAS via cpolar.cn—stability and longevity are unclear
• README is heavy on showcase GIFs and light on architecture details; setup instructions are truncated in the provided source
• Some features (Paper2Video, Paper2Citation) appear thinly documented beyond screenshot demos

Verdict Worth a look if you regularly produce academic presentations or figures and want a starting point you can edit rather than generate from scratch. Skip it if you need battle-tested, well-documented infrastructure or are allergic to web demos hosted on residential connections.