I’d like to request a feature to enhance n8n’s support for PPTX and DOCX file handling. Currently, working with these formats in n8n is somewhat limited, and having built-in nodes or improved functionality to read, modify, and generate DOCX/PPTX files would be extremely valuable for automation workflows.
Has anyone found alternatives for extracting data from ppts? I am currently working with an api in the background to convert ppts to pdf and then trying to use that for my rag setup
Any good options here that doesn’t include another paid subscription? I’m going to try to use MS Graph and have it export pptx as a pdf and then try to extract the text from there. I think MS’s base MS365 subscription includes this ability but I’m not 100% sure on this yet.
I was able to get data from xlsx and docx into a vector data base using existing nodes - it’s cumbersome but doable. Not much out there on how to make that happen - almost exclusively Google Docs…
I build TurboDocx, so flagging the interest up front.
@neme123 on wanting something that isn’t another paid subscription: we maintain an MIT licensed HTML to DOCX converter that’s pure JavaScript. No LibreOffice, no headless Chrome, no service to sign up for. GitHub - TurboDocx/html-to-docx: HTML to DOCX converter · GitHub. It runs in a Code node if you have external modules allowed. That covers generating DOCX for nothing. It does not do PPTX.
Your MS Graph idea is sound for the PPTX side, and probably the cheapest route if you’re already on M365 since you’re only using what you’re paying for. The catch is that you get a PDF back, so you’re extracting text from a rendering rather than reading the source. Speaker notes and anything positioned off-slide won’t survive that.
On the original request, it’s worth splitting in two because the halves are not equally hard. Reading is close to solved: Extract from File covers most cases and mammoth in a Code node handles the rest. Writing splits again, into generating from scratch (which means describing every paragraph in JavaScript) and filling an existing template, which is what most people in threads like this actually want, because the document has to keep a format somebody already approved. A general “DOCX node” is a big surface area. Template filling is narrow and well defined, and it’s the ask that keeps recurring.
PPTX is genuinely the thinner side. Text extraction works the same way as docx. Template-based generation exists but there’s much less of it about.
We offer 5 free / month, TurboDocx does template filling for both docx and pptx: TurboDocx integrations | Workflow automation with n8n. If e-signature comes up, we can do that too. For the free question specifically though, the library above plus your Graph approach is where I’d start.
Disclosure first, same as Nicolas: I build SlideForge.
@Arusa_Khalfay@neme123 - on reading PPTX, we’re not the tool. The Graph-export-to-PDF route works; you lose speaker notes and the reading order gets scrambled on multi-column slides, which is annoying for RAG. If you can load external modules in a Code node, unzipping the pptx and reading the slide XML directly gets you notes and per-shape text without the PDF step.
On writing PPTX, which is the part of the original request still open: pptxgenjs is the
DIY answer. You describe every box and its coordinates in JS. It works, but you’re hand-positioning
text forever, and long labels overflow silently.
We went the other way with SlideForge. You send the content as fields (four metrics, a headline) and get back a .pptx with real text boxes, laid out. There’s a verified community node for it, and the response tells you per field whether it fit or got shrunk, so you can branch on that in n8n instead of
mailing a broken deck.