Best mechanism to create a presentation without it splitting it/truncating

Hi, I’m trying to create PowerPoint presentations for online modules using text from a textbook I have permission to use. When I ask ChatGPT to generate PowerPoint slides, it only creates a limited number (e.g., the first 20 slides), even though I need many more.

A similar issue occurs when I provide all the text upfront—it only generates around 10 slides instead of the 100 I might need. Additionally, maintaining consistent formatting across the slides has been challenging.

I know that Microsoft Word has a “Transform to PowerPoint” function, but I don’t have access to this feature.

I’ve also created a custom GPT to generate Google Apps Script for formatting slides in Google Slides. However, I have to manually copy the text into ChatGPT and then use the generated script in Google Slides. This is tedious, especially given the text is about 15,000 characters, which seems to exceed the program’s limits.

Is there an easier or more efficient way to handle this, such as breaking the content into manageable chunks or automating the process further? Any advice would be appreciated.

2 Likes

Hi @dr.griffith.work

Welcome :people_hugging: to the community!

Creating PowerPoint presentations for large texts, such as textbooks, can be simplified using a structured and efficient approach. A sample how we can do it:

  1. Upload the File (PDF, Word or plain text file)
  2. Break Down the Questions
  3. Combine Files: Once all sections are created, ask GPT-4o to combine them into a single presentation.

I used following prompts:

I require a 100-slide PowerPoint presentation summarizing the content of the book I uploaded. The presentation should comprehensively cover all topics in the book *"Introduction to Human Geography,"* ensuring that no part is omitted.
Here’s how to approach the task:
1. Read and thoroughly understand the book’s content.
2. Divide the presentation into four separate chunks, creating one PowerPoint file for each section:
   - File 1: Slides 1–25
   - File 2: Slides 26–50
   - File 3: Slides 51–75
   - File 4: Slides 76–100
3. Once all sections are completed, combine them into a single PowerPoint presentation file.
Important: Do not display the content of the PowerPoint presentation in this document. Instead, create the PowerPoint files and provide download links for each file. 
###
Start by working on the first section, covering slides 1–25.

Start by working on the second section, covering slides 26–50.

Start by working on the third section, covering slides 51–75.

Start by working on the fourth section, covering slides 76–100.

Combine them into a single PowerPoint presentation file.

Here is my sample chat:

Result:

PS: If you give a break 15-30 minutes, the files were created by ChatGPT can be expired, and you cannot download them. After complete your work, download the files immediately.

My sample chat took approximately 5 minutes.

1 Like

Sorry what? How do you convince any of the models in the model selector to make files and provide links. which tools would I need to toggle for that.

In some sense then any of their responses could be construed as such, and be put to files. and that could keep the vertical conversation space from being polluted by very long answers at any user additional prompt no matter the size. until the user become a prompt artist maybe. I am serious though.

The keyword name of teh tools or environemtn feature that should be availlable to the model, and if the model accept to be the model a user in a plus account would have chosen to have such a feature.

Most of the time the models need to be slapped in the face to accept what the interface pretends they can do… I would like to shorcut that conversationj overhead.

How do you get it to accept creating links. And where does it even keep those temp. files.

also the contex windows limit, the book token size, the max output tokens, and then the segments of the output slides. Did you have to spend a lifetime figuring out the real size or did you accept the outdate GPT4-turbo assumption about what those had to be?