AI image generation for Australian small business: the 30-minute starter

What you’ll have at the end

A working image workflow:

• One image tool subscribed to and configured
• A brand template in Figma or Canva with your fixed text positions and colours
• A saved prompt prefix that produces consistent style
• Your first batch of 4-6 images generated and ready to use

30 minutes once. From there, every visual takes 4-5 minutes.

The honest tool comparison (mid-2026)

Four tools worth knowing for Australian SMB use:

For most Australian SMBs starting fresh in 2026: Midjourney or ChatGPT Plus with DALL-E 4. Pick based on which workflow you’ll use more, Midjourney’s web interface is its own context, while DALL-E inside ChatGPT lets you generate images mid-conversation alongside text work. If you’re already paying for ChatGPT Plus for other work, start there.

Don’t subscribe to all four. The marginal value of the fourth tool is near zero; learn one well first.

Step 1: Pick your one tool (5 minutes)

Sign up for one. Pay for one month. Most subscriptions are AUD-billed or close-to (Midjourney bills USD via Stripe, FX premium of 2-3%).

Once you’ve committed to one for 30 days, do not switch until day 30. Tool-hopping is the most common reason operators don’t get value from image AI.

Step 2: Build your brand template (10 minutes)

This step is the difference between AI thumbnails that work and AI thumbnails that read as “obviously AI”.

Open Figma (free) or Canva (free for the basics). Create a master 1080x1080 template (Instagram square is the default; 1200x1500 for Pinterest, 1080x1920 for Stories). Include:

• Brand colour block in a consistent corner (top-left or bottom-right works for most)
• Text positioning in a fixed location (top half for headline, bottom for CTA, or vice versa)
• Your face / logo position if your visual brand includes them
• Grid / margin guides so every image has consistent breathing room

The template is the brand recognition layer. The AI background is the creative variety layer. Pure-AI loses brand recognition; pure-template loses creative variety. Hybrid wins.

We use this pattern on every On Autopilot visual. Most operators we work with build the template once in 10 minutes and use it for 12+ months.

Step 3: Develop your prompt prefix (10 minutes)

Open your chosen image tool. Generate 5-10 test images, varying only the subject but keeping a consistent style description. The goal: identify the 10-15 style words that consistently produce images aligned with your brand.

Example prompt prefix patterns we’ve seen work for Australian SMB:

Warm editorial cafe:

“natural light, soft shadows, warm earthy colour palette, slight film grain, editorial photography style, shallow depth of field”

Professional services modern:

“clean studio light, neutral background, muted blue and grey palette, sharp focus, business editorial style, minimal composition”

Outdoors trades:

“natural daylight, blue Australian sky, real-world setting, photojournalistic style, honest unposed feel, no artificial polish”

Wellness / allied health:

“soft natural light, calming neutral palette, sage greens and warm whites, soft focus on subject, peaceful unstaged feel”

Test 5-10 generations with your prefix + different subjects. If 7+ out of 10 feel on-brand, your prefix is solid. Save it somewhere you’ll find it again (a notes app, Notion, a prompts.md file).

Important: don’t over-engineer the prefix. 10-15 words is the sweet spot. 50-word prefixes produce worse results (the model gets confused), not better.

Step 4: Generate your first batch (5 minutes)

Now do the real work. Run your prefix + specific subject for 4-6 variants of a real visual you need this week.

Example, for a Melbourne cafe announcing winter menu:

“natural light, soft shadows, warm earthy colour palette, slight film grain, editorial photography style, shallow depth of field, overhead view of a wooden cafe table with two ceramic mugs of hot chocolate, dusted with star anise, and a small plate of mushroom toast on artisan sourdough, autumn leaves falling lightly in soft focus background”

You’ll get 4 variants in Midjourney (default), or 1 (DALL-E default; ask for 4 explicitly). Pick the strongest. Open Figma / Canva. Drop the image into your brand template. Add the text from your social media powerhouse caption (from the social media guide). Export.

That’s the workflow. From this point forward: 4-5 minutes per visual.

What’s it actually replace?

Before this workflow:

After:

For weekly social + occasional blog imagery, AI + template wins on cost-per-image and time. Real photography is still better for hero campaigns, About pages, team shots. Designers are still better for the upload of the quarter / hero campaigns.

The math: at 8 visuals per week, AI saves you 60-150 minutes per week and $200-1000 AUD/month in stock + designer fees. The single $30 AUD/month subscription pays back in week one.

What this workflow is NOT for

Be honest about limits.

• Photorealistic photos of named people. Skin texture, hand details, expressions, AI still misses these in subtle ways readers notice. Use real photography.
• Brand-logo work. Hire a designer. Once. Then use the logo forever.
• Text-heavy designs. AI is still poor at typography in 2026. Use AI for the background; layer text in Figma or Canva.
• High-stakes hero campaigns. The upload of the quarter, the launch image, the website hero, invest in a real designer or real photoshoot. AI is for the weekly-content tier, not the brand-defining tier.
• Anything legally sensitive. Comparative advertising, medical claims, before-after photos with real outcomes, get specialist creative.

For everything else, the 30-minute setup above replaces a significant chunk of small-business design overhead.

The companion playbook

Pair this with the social media powerhouse with Claude Code free guide. Together they produce a week’s worth of content (captions + visuals) in 30-45 minutes once a week, instead of the 5-8 hours most Australian SMBs spend.

Setup both on one Sunday afternoon. From the next Sunday onwards, your weekly content workflow is 30-45 minutes total.

What’s next

• Build a social media powerhouse with Claude Code, the captions layer that pairs with this guide.
• Claude Projects setup in 30 minutes, for the voice file foundation.
• The first 10 prompts every AU SMB owner should run, to test where AI fits in your business.
• Book a free audit if you want help wiring image generation into a wider content workflow.