⏱️ The essentials |
Already have a document with your training material? That's your starting point! The AI will draw on its content to build your course.
You can absolutely use your existing documents. A few best practices can help you get a result that is more relevant and more faithful to what you want to convey.
📄 Start with a suitable document
You can import the following formats: PDF, DOC/DOCX, PPT/PPTX, Pages/Keynote, and ODT/ODP.
Two limits to keep in mind:
100 pages maximum
30 MB maximum
💡 Why 100 pages maximum?
Beyond that, the volume of content risks producing a very long course. In that case, the material is generally better split into two separate courses, organised around the main topics covered.
🧱 A clear structure to help the AI navigate
The structure of your document gives the AI reference points for identifying different topics and understanding how they relate to each other.
Where possible:
use clear, descriptive headings and subheadings;
organise your content into sections and subsections;
group information that covers the same topic.
The hierarchy of your document also gives indications of how the content might be organised:
🔹 A main heading - often a future module
🔹 A subheading - often a concept to develop within the module
💡 You don't need to plan the exact structure of your future course in advance. What matters is that the organisation of your content is easy for the tool to understand.
🖼️ Tables and Images: Best Practices for Successful Extraction
Image extraction from source documents (such as diagrams, illustrative photos, or inspection guides) follows a specific technical pipeline. To prevent your visuals from being dropped or disconnected from their text descriptions, follow these guidelines:
1. Recommendations if you keep a table format
AI reading tools can struggle to extract images embedded inside standard table cells. If you must use a table layout:
Enlarge your images: Images that are too small will be ignored during visual analysis.
Remove vertical borders: Hide or remove the vertical lines of the table (or set them to white/invisible).
Keep horizontal dividing lines: Adding clear horizontal line separators between rows helps the AI correctly map each visual to its corresponding text, preventing mismatches or swapped images.
2. Prefer a vertical, continuous layout
Rather than a complex table or horizontal layout, a standard vertical layout yields the best results. Always place images directly below or above their corresponding text description.
3. A highly effective alternative: Heading-based structure
Instead of creating a table (which can be unstable during AI extraction), structure your content as continuous text with explicit headings:
H1: Main topic (e.g., Fall Arrest Harness - EN 361)
H2: Inspection point / defect type (e.g., Webbing: cuts, abrasion, fraying)
H3: Component name (e.g., Straps and padding)
H4 / Body text: Image, Defect description, Key takeaway...
📐 Mind the Horizontal / Landscape Format (Presentations / PowerPoint)
When importing a horizontal document (landscape layout), the AI processing pipeline generally applies the rule: 1 page / slide = 1 section or module.
What to avoid: Placing two distinct sub-topics or sections side by side on the same horizontal page. This can cause content to end up mixed into the wrong module.
Best practice: Ensure that every new topic or section starts on a new page/slide.
✍️ What should your source document contain?
The principle is simple: your document should contain the knowledge you want to convey to your learners.
Unlike general-purpose generative AI, Didask AI does not look for external information to supplement your content. Key knowledge must therefore be present in your document.
Make sure to include:
essential concepts and information;
necessary explanations and definitions;
rules, methods, or procedures to remember;
concrete examples;
tables or diagrams useful for understanding.
explanatory diagrams or photos (if needed).
🖥️ What if your document is very concise?
This is often the case with PowerPoint presentations: a few keywords or key ideas appear on the slides, while explanations are delivered verbally.
You can absolutely start from this type of document. Just keep in mind that the AI can only work with the information it contains. If certain explanations are essential for learning, make sure to add them to the document.
💡 A useful test: could someone new to the subject understand the main takeaways by reading your document alone?
🧹 What if it contains superfluous information?
Conversely, the AI treats all information in your document as potential material for building the course. Content that is not meant to be learned may end up developed into a module.
Before generating, remove elements that are not useful for learning, for example:
cover pages with no educational content;
automatic tables of contents;
administrative or legal notices;
information intended only for the trainer or designer;
purely decorative visuals.
🎯 A simple question to help you decide: "Do I want the learner to learn this information?"
Yes - keep it.
No - it's better to remove it from the source document.
🎓 What about exercises?
Your source document should primarily contain the knowledge to be conveyed. You do not need to prepare exercises in advance: the tool generates them from the knowledge in your document and the learning objectives you define.
If your document already contains exercises, they will not necessarily be reproduced as-is. They may be interpreted as information to convey, or transposed differently during generation.
💡 Our advice: rather than writing out exercises, focus on the knowledge, rules, or best practices the learner will need to apply. The AI can then use that material to create appropriate exercises.
✅ One last check?
Before launching your generation, three questions can help you get the most out of your document:
Are the different topics easy to identify?
Is the knowledge you want to convey clearly present and sufficiently explicit?
Have you removed information that is not meant to be learned?
Are key images large enough and positioned right next to their text explanations?
If using a table with images, have I removed vertical borders while keeping clear horizontal dividers?
For horizontal/PowerPoint formats, does each new section start on a new page?
These simple checks will help the AI better understand your material and generate a course that matches your expectations. 🚀
💡 To get an even more targeted result, make sure to fill in the "Context" field when generating. See how to write it effectively.
Keywords: AI generation, source document, document import, PDF, PowerPoint, DOCX, document structure, educational content, course, training, best practices