Anyone here automating PowerPoint slide creation with LLMs?
I’m building a pipeline that tales structured data (financials) and populates PPTX templates. If I dont template the output becomes too varied.
Hit a few walls: chart XML manipulation breaks easily, replace data resets formatting, think-cell charts are a night mare to strcipt againts, not following rules.
Curious what works for others:
\- best library approach for chart data updates without losing styling?
\- Anyone using a templating layer instead of direct python-pptx?
\- For multi-step generation (data > narrative > slide), are you orchestrating with agents and skills or just sequential scripts?
\- Any pitfalls I should know about before scaling slide creation?
Mostly only python-pptx today. Open to switching tools if there’s something better.
Thanks! Really appreciate all the help!