A Hands-On Midjourney Walkthrough
Generate professional AI art for thumbnails, blog headers, and social media — from prompt crafting to upscaling and commercial use.
A Hands-On Midjourney Walkthrough
Midjourney is the highest-quality AI image generator for artistic and commercial work. This walkthrough covers writing effective prompts, navigating the Discord interface, iterating on results, and producing assets ready for thumbnails and social media.
Step 1: Access Midjourney via Discord
Midjourney runs entirely through Discord:
- Go to midjourney.com and sign up for a subscription (Basic: $10/mo, Standard: $30/mo — Standard includes unlimited relaxed-mode generations)
- Join the Midjourney Discord server (link from your account page)
- Navigate to any
#newcomer-roomsor#generalchannel - Type
/subscribeto confirm your plan is active
You can also use Midjourney via its web app (alpha.midjourney.com) if you prefer a visual interface over Discord commands. Both produce identical results.
Step 2: Write Your First Prompt
Type /imagine prompt: followed by your description. The prompt structure that consistently produces good results:
/imagine prompt: [subject], [style], [lighting], [composition], [mood] --ar 16:9 --v 6
Example for a YouTube thumbnail background:
/imagine prompt: futuristic cityscape at sunset, cyberpunk aesthetic, neon rim lighting, wide-angle composition, cinematic mood --ar 16:9 --v 6
Prompt anatomy that works:
- Subject: What’s in the image (be specific — “a red sports car on a wet mountain road” beats “a car”)
- Style: Art direction (“photorealistic”, “oil painting”, “3D render”, “flat illustration”, “cinematic”)
- Lighting: “golden hour”, “studio lighting”, “neon rim light”, “soft diffused”
- Composition: “wide-angle”, “close-up”, “top-down”, “rule of thirds”
- Parameters:
--arfor aspect ratio (16:9 for thumbnails, 9:16 for stories, 1:1 for posts),--v 6for latest model
Midjourney generates 4 variations in ~60 seconds.
Step 3: Iterate with Variations and Upscaling
After your 4-image grid appears, use the buttons below:
- U1-U4: Upscale one image to higher resolution (makes it larger and more detailed)
- V1-V4: Create 4 new variations of that specific image (same style, slightly different)
- 🔄: Reroll — generate 4 entirely new images from the same prompt
Iteration strategy:
- First generation: broad prompt, pick the closest image (V1-V4)
- Second generation: refine with a more specific prompt based on what worked
- Third generation: upscale the winner (U1-U4), then use Vary (Subtle) for fine-tuning
Most production-ready images take 2-3 rounds. Don’t expect perfection on the first try.
Step 4: Master Stylization Parameters
Parameters that dramatically change output quality:
--s 250(stylize): Higher values = more artistic interpretation (range 0-1000, default 100). Use 250-400 for thumbnails that pop.--chaos 20(chaos): Higher = more varied results (range 0-100). Use 10-30 for controlled variation, 50+ for wild exploration.--no text(negative prompt): Exclude unwanted elements.--no text, watermark, signatureis essential for clean commercial use.--style raw: Reduces Midjourney’s default aesthetic enhancement. Use for photorealistic work where you want literal prompt adherence.--seed 12345: Locks the random seed for reproducible results. Use when you want to iterate on the same base image.
Example refined prompt:
/imagine prompt: close-up portrait of a confident woman entrepreneur, photorealistic, soft studio lighting, shallow depth of field, professional headshot --ar 1:1 --s 300 --style raw --no text --v 6
Step 5: Create a Thumbnail Composition
Midjourney generates backgrounds and subjects, but thumbnail composition needs text overlays:
- Generate your background image with
--ar 16:9 - Upscale the best result
- Download the image (click the image → right-click → Save Image As)
- Open in Canva or Photoshop
- Add bold text overlay (3-5 words, high contrast, thick font)
- Add your face photo if applicable (cutout on top of the AI background)
What makes thumbnails clickable: high contrast, one focal point, emotion (surprise/curiosity), text that creates an open loop (“This changes everything” works; “Review of Product X” doesn’t).
Step 6: Generate Consistent Brand Assets
For a cohesive visual brand across content:
- Lock a style keyword: Pick one style descriptor and reuse it (e.g., “minimalist flat illustration”, “cinematic 3D render”, “watercolor texture”)
- Lock a color palette: Include specific colors in every prompt (“emerald green and deep navy palette”)
- Use
--sref(style reference): Upload a reference image and use--sref [URL]to apply its style to new generations. This is the strongest consistency tool. - Save successful prompts: Keep a document of prompts that produced great results. Reuse the structure with subject swaps.
Example for consistent blog header illustrations:
/imagine prompt: abstract geometric pattern, emerald green and navy blue, minimalist flat illustration, blog header style --ar 3:1 --s 200 --sref [previous-successful-image-url] --v 6
Step 7: Understand Commercial Usage Rights
Midjourney’s terms for paid subscribers:
- Paid plans: You own the assets you generate. You can use them commercially (thumbnails, social media, products, marketing materials).
- Free trial: Cannot be used commercially.
- Important caveat: Midjourney doesn’t guarantee your generated image doesn’t infringe on existing copyrighted works. For logos and brand identity, use AI art as inspiration, not final deliverable.
For content creators (thumbnails, blog headers, social posts), paid Midjourney assets are safe to use. For trademark-adjacent uses (logos, product packaging), consult legal.
What This Workflow Produces
| Asset | Time | Output |
|---|---|---|
| 4 initial images | ~60 sec | Grid of 4 variations |
| Refined iteration | ~60 sec | 4 improved variations |
| Final upscaled image | ~30 sec | High-res single image |
| Thumbnail with text overlay | ~10 min (in Canva) | Ready-to-publish thumbnail |
| Total per asset | ~15 min | Production-ready visual |
Common Mistakes to Avoid
- Overloaded prompts — “beautiful amazing stunning incredible detailed masterpiece 8k” doesn’t help. Midjourney ignores quality spam. Describe what you want, not how good it should be.
- Ignoring aspect ratio — generating 1:1 then cropping to 16:9 loses composition. Always set
--arto your target format. - Not using
--no text— Midjourney v6 sometimes generates fake text in images.--no textprevents this. - Expecting photorealism from default settings — use
--style rawfor photographic work. The default adds an artistic filter. - Generating once and settling — the first grid is rarely the best. Budget 2-3 iterations for production work.
Pro Tips
- Keep a prompt library — save prompts that work in a text file or Notion database. Tag by use case (thumbnail, illustration, pattern, portrait).
- Use
/describe— upload an image you like and Midjourney writes a prompt to recreate its style. Reverse-engineering references is faster than guessing. - Batch by format — generate all your thumbnails for the month in one session. Switching contexts (different projects) wastes time.
- Use Vary (Region) — upscale an image, then select a specific area to regenerate. Fix one bad element without redoing the whole image.
- Check community prompts — browse the Midjourney gallery (midjourney.com/explorer) for prompts that produce styles you want. Copy and adapt.