Documents & media · Intermediate
Work with documents and images
Ask focused questions about a PDF and write image briefs with a clear subject, composition and style.
3 lessons · About 30 minutes including practice · Free to read
What you will learn
- Ask questions that can be answered from an uploaded PDF.
- Verify a document answer against the original.
- Refine an image by changing one part of the brief at a time.
Lesson 1 of 3
Ask from a document
Document AI uses an uploaded PDF as a source for questions. In Nevatal, this tool requires a Google Gemini API key. Use a small PDF you can inspect while you learn the workflow, such as a fictional workshop brief.
Upload the PDF and check the document list before asking a focused question. Document AI searches across every indexed document listed above the chat, so name your practice file in the question if you have other documents. If the sources do not contain the answer, ask the model to say so rather than fill the gap from general knowledge.
Worked example
Create a one-page PDF containing these practice facts: The workshop has 12 places. Registration closes on 8 October. Lee manages the waiting list. No venue has been selected. Question: Who manages the waiting list, and what is the registration deadline? Use only this document.
Try it yourself
Save the practice facts as workshop-practice.pdf, upload it, and ask the question with the filename included. Open the original PDF beside the answer and find the sentences that support it.
Check your result
- The answer identifies Lee and 8 October.
- You can locate both facts in the original PDF.
- The response does not mix in facts from unrelated indexed documents.
Lesson 2 of 3
Find the limits of an answer
Document search retrieves relevant passages; it does not guarantee that every answer fully represents the document. Ask about one topic at a time and check important details directly in the PDF. If text in a scanned or complex page cannot be read reliably, do not treat a confident answer as evidence that it was extracted correctly.
Try a question whose answer is absent. This is a useful check of whether your instructions are being followed. Keep documents organized and remove a practice upload from Document AI when you no longer need its saved index.
Worked example
What is the workshop venue? If the document does not specify a venue, say that it has not been selected. Do not suggest a location.
Try it yourself
Ask the venue question about your practice PDF, then ask for the number of places. Compare the unsupported question with the one the document can answer.
Check your result
- The response does not invent a venue.
- The supported answer says 12 places.
- You verify the source rather than relying only on a quoted passage in the answer.
Lesson 3 of 3
Write an image brief
Image Generation also requires a Google Gemini key. Describe the subject, setting, composition and lighting in the prompt, then choose the Style and Framing controls. Mention space for a headline if the image will be used in a layout.
Review the result against the brief and change one part of the prompt at a time. Look closely at details such as hands, objects and lettering. Add exact text in a design tool when the generated image does not reproduce it reliably.
Worked example
A blue commuter bicycle leaning against a pale brick wall beside a small potted plant. Soft morning light, simple background. Place the bicycle in the right half and leave clear space on the left for a headline. No text or logos.
Try it yourself
Generate the image, then make a second version that changes only the lighting to an overcast afternoon. Compare composition and object details as well as mood.
Check your result
- The bicycle is on the right with usable space on the left.
- The image has no unwanted lettering or logos.
- The two briefs differ only in the lighting instruction.
Put it into practice
Repeat one exercise with your own material. Keep the original beside the result, use the lesson’s checklist and revise the input if something is missing.