sn-ppt-entry

To generate a PPT from a PDF, you upload your file and specify presentation parameters like role, audience, scene, and page count. The system parses your PDF into normalized artifacts, creates a document digest, and prepares an info pack for downstream slide generation.

Yes, you can create slides from Word documents and Markdown files. The document parsing feature accepts pdf, docx, md, and txt formats, extracting text, tables, and images into normalized JSON to prepare the deck task pack.

Extracting images and tables for slide generation involves parsing uploaded files into normalized JSON while preserving table and image indices. The system also generates VLM image captions and caches them to ensure reliable layout creation downstream.

Yes, slide generation preserves numbers by using LLM-based document digest generation with numeric-preservation constraints. This ensures that data from your uploaded documents does not drift during the transition to the normalized artifacts.

The best way to start a PPT project entry is to send a prompt with your desired mode, role, audience, scene, page count, and upload your reference files. The system routes the task to creative or standard deck generation based on your specified mode.

No, you do not need to manually caption images. The system automatically generates VLM captions for inherited or standalone images once, caches them, and ensures that subsequent page generation reuses these captions reliably for layout planning.