2. Generating and editing images

AI image generation is the ability of a computer program to create a brand-new visual image from a written description — called a prompt. You type "a smiling Ethiopian coffee farmer at sunrise in the green highlands" and the AI produces a realistic photograph or illustration within seconds.

This is not a simple search for an existing photo. The AI actually creates pixels from scratch, guided by the patterns it learned from hundreds of millions of images during its training. The result is an entirely new image that has never existed before.

Popular AI image tools include:

• Midjourney — known for high artistic quality.

• DALL·E (built into ChatGPT) — easy to use; good for business graphics.

• Adobe Firefly — integrated with Adobe tools; designed with commercial licensing in mind.

• Canva AI (Dream Lab) — beginner-friendly; popular for social-media graphics.

• Stable Diffusion — open source; can be run on your own computer.

The single most important skill for AI image generation is writing a good prompt. A vague prompt gives a vague result; a specific, structured prompt gives a precise, usable image.

A strong image prompt usually has four parts:

1. SUBJECT — who or what is in the image.

Example: "a young Ethiopian woman named Sara selling fresh injera at a market stall"

2. STYLE / MEDIUM — the visual look you want.

Example: "professional food photography style", "flat illustration", "watercolour painting", "realistic photo"

3. SETTING / CONTEXT — where and when.

Example: "in Addis Ababa Mercato, morning light", "on a wooden table with coffee beans"

4. TECHNICAL DETAILS — quality or format hints.

Example: "high resolution, 4K, soft natural lighting", "portrait orientation", "white background for product photo"

Putting it together:

"A young Ethiopian woman smiling and selling fresh injera at a market stall in Mercato, Addis Ababa, morning golden light, professional food photography style, high resolution."

That one sentence will produce a far better result than just writing "Ethiopian woman injera."

Generating an image is just the beginning. Modern AI tools also let you edit images using natural language — no Photoshop skills needed.

Common AI image-editing operations:

• INPAINTING — paint over a part of the image and ask the AI to replace it. Example: You generated a product photo but the background is too busy. You mask the background and type "replace with a clean white studio background" — done.

• OUTPAINTING — extend an image beyond its borders. Example: You have a portrait but need a wider banner. Ask the AI to "extend the background to the left and right" and it fills in new, consistent content.

• STYLE TRANSFER — take an existing photo and apply a new artistic style. Example: Upload a phone photo of your shop and ask the AI to "make this look like a professional architectural illustration."

• BACKGROUND REMOVAL — automatically remove the background from a product photo, leaving only the item on a transparent or custom background. Canva AI and Adobe Firefly both do this with one click.

• UPSCALING — increase the resolution of a low-quality image so it looks sharp when printed or used as a banner.

These tools are available in ChatGPT (DALL·E), Adobe Firefly, Canva, and several free online editors.

AI image generation raises important ethical and legal questions you must understand before using AI images for any business purpose.

COPYRIGHT AND OWNERSHIP

The legal status of AI-generated images is still evolving worldwide. In general:

• Images you generate with a paid tool (like Midjourney, Adobe Firefly, or Canva Pro) under their terms of service can usually be used commercially.

• Always read the tool's Terms of Service — some free tools reserve rights to your generated images.

• Adobe Firefly is specifically designed with commercial use in mind; it trains only on licensed images.

• Never claim an AI-generated image as your own original photograph or artwork without disclosing it was AI-made.

AVOIDING HARMFUL AND DECEPTIVE IMAGES

• Do not generate realistic images of real, identifiable people without their consent — especially public figures or politicians.

• Do not create images designed to mislead viewers into thinking something happened that did not (deepfakes).

• Avoid generating content that reinforces harmful stereotypes — for example, always showing African people in poverty or conflict.

BRAND SAFETY

• Be consistent: use a defined colour palette, logo placement, and style so that AI images feel like part of your brand, not random visuals.

• Check every AI image carefully before publishing — AI sometimes makes visual mistakes (extra fingers, garbled text, unrealistic proportions).

• Add your logo or watermark to AI images before sharing them on social media.

Here is a step-by-step workflow that any Ethiopian small-business owner can follow to go from an idea to a published AI image:

STEP 1 — DEFINE YOUR GOAL

Ask yourself: What is this image for? (Instagram post, product photo, banner, menu illustration?) Who will see it? What feeling should it create?

STEP 2 — CHOOSE YOUR TOOL

• For quick social-media graphics: Canva AI (free tier available).

• For high-quality illustrations: DALL·E inside ChatGPT (free and paid).

• For commercial product photos: Adobe Firefly (paid, but safe for commercial use).

STEP 3 — WRITE AND REFINE YOUR PROMPT

Use the four-part structure: Subject + Style + Setting + Technical details. Generate 3–5 variations. Pick the best one.

STEP 4 — EDIT IF NEEDED

Use inpainting, background removal, or upscaling to refine the chosen image. Add text, your logo, or brand colours using Canva or the tool's built-in editor.

STEP 5 — REVIEW AND PUBLISH

Check for visual errors, confirm licensing, add watermark/logo, then post.

For Almaz's Meskel catering post, this entire workflow takes about 15 minutes — instead of the 2–3 days it would take to hire a designer and photographer.