1 Course position and tonight’s promise

Day 4 opens Module 2. Tonight you will write a safe image prompt for a fictional housewarming invitation, then use editing tools to restore and lightly colourise a fictional old family photo, checking rights and watermarks before saving either result to your challenge portfolio.

2 What you will learn and why it matters

• Describe an image prompt's subject, style, setting, and constraints in plain language.
• Choose an aspect ratio and invitation type that match how guests will see it.
• Type reliable names and dates yourself instead of trusting AI-rendered text.
• Use inpainting and outpainting to remove, replace, or expand image content deliberately.
• Restore and colourise an old photo while treating the result as a labelled guess, not a fact.
• Check commercial rights, watermarks, and consent before sharing or printing any image.

An image prompt shapes what a tool attempts to draw. It cannot guarantee accurate text, real identities, or usage rights—you verify all three.

3 A familiar situation: diagnose before you prompt again

Fictional sample values: Priya is planning a housewarming and wants an invitation, plus a restored photo of her grandmother for the party table. Her first three attempts each have a different visible problem. Predict the missing fix before revealing it; do not use real addresses, phone numbers, or someone else's photo without permission.

4 Simple explanation, then accurate vocabulary

An image prompt is a written description that tells a tool what to draw, edit, or restore. Composition describes how subjects and space sit in the frame; aspect ratio is the image's width-to-height shape. Inpainting changes a chosen area; outpainting extends the canvas beyond the original edges.

5 Image prompt anatomy

A dependable image prompt names four things: the subject (who or what), the style (watercolour, photo-real, papercut), the setting (indoors, garden, plain background), and constraints (no readable text, no real faces, safe colours). Missing any one of these makes the result harder to predict.

6 Visual styles and aspect ratios

Style sets the visual mood; aspect ratio sets the shape for its destination—square for a social post, portrait for a printed card, landscape for a banner. Choosing the wrong ratio forces an awkward crop later, so decide before generating.

7 Invitation types

An invitation's type shapes its content, not just its picture: a printed card needs bleed-safe margins and legible fonts; a digital share needs a shorter headline and mobile-safe text size; a video teaser (covered next lesson) needs motion-safe framing. Pick the type before writing the prompt.

8 Reliable names and dates

Image tools often misspell rendered text or invent it entirely. Never trust a name, date, time, or address drawn inside an image. Generate the artwork without text, then add real facts yourself in a document or design tool where you can proofread every character.

Generate a housewarming scene: warm indoor party, string lights, soft watercolour style.
Constraint: no readable text, no invented names, no logos.
I will add the event name, date, time, and address separately and proofread them myself.

9 Family-photo greetings

Turning a real family photo into a greeting card involves a real person's likeness. Ask permission from an adult in the photo (or a guardian for a child), avoid uploading images of people who have not agreed, and keep the edit close to the original rather than inventing a new face or scene around them.

• Get permission before editing or sharing someone else's photo.
• State what may change (background, colour) and what must not (facial features, expression).
• Keep a copy of the original alongside the edited greeting.
• Do not caption or tag a real person without their consent.

10 Background/object removal and expansion

Inpainting removes or replaces a selected area—useful for a cluttered background or an unwanted object. Outpainting extends the canvas beyond the photo's original edges, which means every new pixel outside the original frame is invented. Inspect expanded regions closely; they can add objects, furniture, or people that were never there.

11 Photo restoration and colourisation

Restoration removes scratches, tears, and fading; colourisation adds plausible colour to a black-and-white photo. Both are educated guesses trained on patterns, not a record of the true colours or repaired detail. Keep the original file, restore a copy, and ask a family member who remembers the photo to confirm details such as clothing or wall colour before treating the result as final.

12 Rights, watermarks, and verification

Before printing or sharing any generated or edited image, confirm what the tool's current terms allow: personal use, commercial use, watermark removal, and ownership of the output. A free tier may add a watermark or restrict commercial use; a small-business greeting card is a commercial use and needs its own rights check, separate from a personal housewarming invitation.

13 Interactive process model: prompt, edit, verify

The event-facts card is the learner's real input; the prompt-builder is the anatomy decision; the draft card is the tool's attempt; the text-and-likeness gate is where rendered text and faces are verified; the editor-correction stage is where inpainting/outpainting fixes issues; the last card is the exported invitation. This simplified picture does not expose hidden model reasoning.

14 Facilitator-led live demonstration: one ratio mismatch

Start with the fictional square-ratio invitation that crops the centrepiece. Intentionally keep the square ratio because it “looked fine in the small preview.” Show why that is a mismatch: a print-ready card needs portrait framing, not a preview-sized guess. Diagnose the crop and switch the aspect ratio and prompt together.

• Use the tool's current New image or Edit equivalent with a fictional prompt.
• Generate the square-ratio draft and identify the cropped centrepiece.
• Rebuild the prompt with a portrait ratio and the same style and constraints.
• Compare the new draft against the card's actual print dimensions.
• Save the before/after and state why the ratio decision changed.

15 Your Invitation & Restoration Checklist

Save this as “Day 4 — Invitation & Restoration Checklist.” For each fictional item, keep the prompt, the verification evidence, and a one-line reflection.

16 Learner lab: build your own safe invitation and restoration

1. Prepare one fictional event's facts and one non-sensitive practice photo (your own or a clearly licensed sample).
2. Predict the aspect ratio and style the destination needs before generating.
3. Write a prompt naming subject, style, setting, and constraints; generate a first draft.
4. Use inpainting or outpainting only where needed, and inspect the edited region closely.
5. Add real names, dates, and addresses yourself; never trust rendered text.
6. Check watermark and commercial-rights terms for your intended use; export/save and add a one-line reflection.

17 Behind the scenes, limitations, safety, and verification

• Rendered text inside an image is frequently wrong; never trust a drawn name, date, or address.
• Restoration and colourisation are plausible guesses, not a record of the true original.
• Outpainting invents every pixel beyond the original frame; it can add objects that were never there.
• A generated likeness of a real person is not that person's actual current appearance or consent.
• Free-tier tools may add watermarks or restrict commercial use.
• Style and quality vary by tool, prompt, and version; today's result may not repeat tomorrow.

18 Common confusion, errors, troubleshooting, and best practices

• Best practice: name subject, style, setting, and constraints every time.
• Best practice: keep the unedited original file next to every restoration.
• Trainer tip: ask “What in this image was verified by me, and what was only generated?”

19 Takeaways

• Build every image prompt from subject, style, setting, and constraints.
• Choose the aspect ratio and invitation type for the real destination before generating.
• Never trust AI-rendered names, dates, or addresses—type and proofread them yourself.
• Treat restoration and colourisation as labelled estimates, and get consent for real faces.
• Check watermark and commercial-rights terms before printing, selling, or sharing.

Module 2 continues when you can explain what your prompt controlled, what you verified, and what rights you checked.

20 What comes next

Next is Topic 05—Produce a 30-Second Video Invitation. You will carry this invitation's verified facts and style choices into a short video, checking motion-safe framing and the same rights and consent rules for moving images.

• Reopen your Invitation & Restoration Checklist.
• Keep your verified event facts and chosen style ready.
• Note one motion or timing question you predict the video lesson will need to answer.