AI Art

AI-powered visual generation for marketers and agencies. Create scroll-stopping ad creatives, social media graphics, and brand visuals in...

Location hidden

•Created by

Hossameldin Ahmed

1 joined

Home

Products Reviews

Hossameldin Ahmed

@heage1·Sep 19

How a DTC brand cut creative costs by 94% with AI visuals

A mid-size DTC skincare brand came to us spending $4,800/month on ad creative. Studio days, retouching, and stock licenses. After 90 days on an AI-first workflow, that number was $290.

That is a 94% cut. ROAS did not fall. It went up.

Before vs after

Line item	Old stack	AI stack
Studio / photographer	$2,200	$0
Designer hours	$1,800	$180
Stock licenses	$800	$0
AI tools + prompts	$0	$110
Monthly total	$4,800	$290
Creatives shipped / month	12	64
Blended ROAS	2.3x	2.7x

What they actually changed

They did not fire design. They stopped using design for first drafts.

Hero SKU library: 8 clean product cutouts as the source of truth
Prompt batching: 10 scene variations per SKU every Monday
Human pass only on winners: designer polishes the top 10%, not all 64
Fatigue rule: any ad past 8 days gets a new visual, same copy

The unlock was volume, not cheaper art

At 12 creatives a month, fatigue hit every 10 days and CPMs crept up. At 64, they could kill losers in 48 hours and keep spend on fresh winners. Creative cost dropped and media efficiency improved.

What did not work

Generating people holding the product (inconsistent faces, brand risk)
Letting AI invent packaging text (spelling errors in ads)
Skipping a brand palette. Early outputs looked like 64 different companies.

Fix: isolated product + locked colors + overlay real type.

The takeaway

If creative is more than ~8% of ad spend, you are probably overpaying for drafts. Use AI for the 80% that gets tested. Save humans for the 20% that already won.

What are you paying per creative right now? Drop a number and I’ll tell you if AI would actually move it 👇

Hossameldin Ahmed

@heage1·Sep 19

The color psychology cheat sheet for AI ad creatives

Color is the fastest conversion lever in an AI creative. People decide how they feel about an ad before they read a word. If you leave color to the model’s default, you get random mood and random results.

The cheat sheet

Color	What it signals	Use it when you sell
Red	Urgency, appetite, heat	Flash sales, food, energy
Blue	Trust, calm, competence	SaaS, finance, health
Green	Growth, ease, “natural”	Wellness, money, eco
Gold / cream	Luxury, warmth, status	Premium, beauty, jewelry
Black	Power, exclusivity	High-ticket, fashion
Orange	Play, value, action	Impulse offers, CTAs

How to actually prompt it

Do not write “make it pop.” Name 2–3 colors and a mood:

“[PRODUCT], color palette of deep navy and warm gold, luxury and trustworthy mood, muted saturation, editorial advertising photography, 4K”

Then lock it:

Negative: “neon, oversaturated, rainbow, clashing colors, stock photo look”

3 rules that keep palettes converting

One dominant, one accent. 70% base color, 20% support, 10% CTA color.
Match the offer, not your logo, on cold ads. Brand colors are for remarketing. Cold traffic needs the emotion of the offer.
Desaturate 10%. AI defaults too vivid. “Muted, filmic color grading” almost always lifts perceived quality.

Quick test

Take your last winning ad. Keep the composition. Only change the palette:

Blue/navy for trust
Red/black for urgency
Cream/gold for premium

In our internal tests, palette-only swaps moved CTR by 18–40% with zero copy changes. Color is not decoration. It is positioning.

Drop your product + the feeling you want. I’ll give you the 3-color palette to prompt 👇

Hossameldin Ahmed

@heage1·Sep 19

Copy this: The exact prompt for luxury product flat lays

Luxury flat lays fail when they look like a Pinterest craft table. The camera is too busy, the props are cheap, and the product gets lost. Use this prompt as-is, then swap only the bracketed parts.

The exact prompt

Overhead luxury flat lay of [PRODUCT], arranged on [SURFACE], [2–3 PREMIUM PROPS] placed with generous negative space, single soft window light from the left, long elegant shadows, muted [COLOR 1] and [COLOR 2] palette, tactile textures, editorial still life, shot on 50mm, shallow depth at the edges, ultra realistic, 8K

Negative:

clutter, busy composition, plastic props, harsh flash, centered grid of objects, oversaturated, watermark, text, cartoon, stock photo look

Fill-in examples

Fragrance

Product: “a heavy glass perfume bottle” Surface: “dark honed marble” Props: “a cut orchid and a silk ribbon” Colors: “cream and antique gold”

Fashion accessory

Product: “a grained leather belt, coiled once” Surface: “walnut wood” Props: “brass keys and a linen napkin” Colors: “cognac and charcoal”

Skincare

Product: “a frosted serum bottle” Surface: “wet stone” Props: “a single green leaf with water droplets” Colors: “sage and soft white”

5 rules so it reads expensive

Max 3 props. More than that looks like a gift basket.
Name the surface. Marble, walnut, linen, wet stone. Surfaces do the luxury work.
One light, from the side. Overhead flash kills depth.
Leave 40% of the frame empty. Negative space = price.
Describe texture, not adjectives. “Grained leather” beats “luxury leather.”

Why overhead works in ads

Flat lays crop cleanly to 1:1 and 4:5, and the product stays recognizable at thumbnail size. In our tests, luxury flat lays beat 3/4-angle pack shots by +22% CTR on cold Meta prospecting for beauty and accessories.

Copy it. Generate 4. Pick the quietest one.

Comment your product and I’ll fill the brackets for you 👇

Hossameldin Ahmed

@heage1·Sep 18

The email header image test: AI vs. stock (open rates inside)

We ran a 6-week email test on the same list, same subject lines, same send times. The only variable was the header image.

Half the sends used licensed stock. Half used custom AI headers matched to the email’s offer.

The results

Metric	Stock header	AI header	Lift
Open rate	21.4%	24.1%	+12.6%
Click-to-open	8.2%	11.9%	+45%
Unsubscribe rate	0.31%	0.22%	-29%
Cost per header	$49–$120	~$0	-99%

Opens moved. Clicks moved more. That is the tell: people opened both, but they believed the AI headers belonged to the brand.

Why stock lost

Stock headers fail for three boring reasons:

The reader has seen the photo. Familiar imagery gets skipped like banner ads.
The scene does not match the offer. A generic laptop-and-coffee shot sells nothing specific.
Every competitor can buy the same file. You are paying to look interchangeable.

Why AI headers won

The winning AI versions were not “more artistic.” They were more relevant:

Same product as the email
Same color palette as the brand
Same lighting mood as the landing page the click went to

That continuity made the email feel like one campaign, not a newsletter with a rented photo on top.

The header formula we used

“[PRODUCT or scene] as a wide cinematic email banner, subject on the left third, large negative space on the right for a headline, brand colors, editorial lighting, no fake text, 16:9, 4K”

Then we overlaid the real headline in the email builder. Never let the model write the words.

Practical takeaway

If your emails are “fine” and not growing, swap the header before you rewrite the copy. In this test, one image change beat most copy tweaks we had tried that quarter.

Want the exact 3 header prompts from the winning sends? Comment your niche and I’ll adapt them 👇

Hossameldin Ahmed

@heage1·Sep 18

The split-screen prompt trick for comparison-style ads

Comparison ads stop thumbs because the brain is forced to pick a side. Most AI split-screens fail because both halves look equally pretty. Pretty vs. pretty does not sell. Ugly vs. clearly better does.

The prompt skeleton

“Split-screen advertisement, vertical divide down the center, LEFT side: [BEFORE STATE], dull lighting, messy, desaturated, labeled BEFORE. RIGHT side: [AFTER STATE], cinematic lighting, clean, rich color, labeled AFTER. Same camera angle both sides, high contrast, 16:9, 4K”

The two non-negotiables:

Same camera angle so it feels like one scene transformed
Unequal quality on purpose so the right side wins instantly

Fill-ins that convert

Product quality

Left: “cheap plastic version on a cluttered desk, harsh overhead light” Right: “premium version on a clean walnut surface, soft window light”

Workflow

Left: “stressed designer at a messy board, papers everywhere” Right: “same person generating ads on a laptop, calm, organized, glowing screen”

Results

Left: “empty analytics dashboard, gray, flat” Right: “up-and-to-the-right chart, green glow, same dashboard restyled”

Why this format works

Element	What it does
Hard center divide	Creates an instant “choose” moment
Desaturated left	Makes the problem feel real
Rich right side	Makes the solution feel obvious
Same angle	Stops the brain from dismissing it as two unrelated photos

In our tests, split-screen creatives averaged 2.2% CTR vs. 1.3% for single-hero product shots on cold traffic.

3 mistakes that kill the ad

Both sides equally colorful
Different products on each side (confuses the comparison)
Tiny labels the phone crop cuts off. Keep BEFORE / AFTER large and inside the middle 70% of the frame.

60-second workflow

Write the left (pain) in 8 words
Write the right (desired state) in 8 words
Paste into the skeleton
Generate 4 versions, pick the one where the left looks worse on purpose

Drop your product and I’ll write the left/right split for you 👇

Hossameldin Ahmed

@heage1·Sep 18

The font pairing guide for AI-generated social media graphics

AI is getting better at images. It is still terrible at picking type. If your social graphics look “off,” the fonts are usually why.

The 3-pair system that always works

You only need one display font + one workhorse font. Mixing more than two is how AI graphics start looking like a flyer from 2009.

Pair	Use it for	Vibe
Bold geometric sans + clean grotesque	Product ads, SaaS, ecom	Modern, high converting
Elegant serif + simple sans	Beauty, luxury, editorial	Premium, slower scroll
Condensed display + regular sans	Sales, launches, urgency	Loud without chaos

Prompt words that actually control type

Do not write “nice fonts.” AI ignores that. Be specific:

“headline in bold geometric sans-serif, body in clean grotesque, high contrast, large readable type, no decorative script, generous letter spacing”

Add this every time:

Negative: “cursive, script, comic fonts, warped letters, misspelled text, tiny unreadable type”

Size rules for social

Headline: 20% of the frame height minimum. If you have to squint on mobile, it is too small.
One message per graphic. AI will try to fill every gap with extra words. Stop it.
Leave 15% padding on all edges. Platform UI eats the corners.

The 80/20 shortcut

If you are not a designer, default to:

“Inter or Helvetica-style sans for all text, extra bold headline, regular body, black on cream or white on navy”

That single pairing outperforms fancy type in most feed tests because legibility beats personality at thumbnail size.

Common AI type fails

Fake letters that almost spell the word
Different fonts on every line
Text sitting on a busy texture with no contrast

Fix: generate the image without text, then overlay real type in your editor. Use AI for the visual. Use real fonts for the words.

What pairing are you using right now? Drop a screenshot and I’ll tell you if it is working or fighting the image 👇

Hossameldin Ahmed

@heage1·Sep 17

How AI visuals helped us scale from $5K to $50K/month ad spend

Scaling spend is easy. Scaling creative volume is what actually lets spend go up without killing ROAS.

Here's the path from $5K to $50K/month in ad spend — and the role AI visuals played.

The bottleneck at $5K

At $5K/month we were running 6–8 creatives. Fatigue hit every 10–14 days. ROAS would dip, we'd pause, wait on a designer, and restart.

Classic stall:

📉 Creative refresh rate: 2–3 new ads / week
📉 Time to first test: 3–5 days
📉 Winning ad rate: ~1 in 8

You cannot 10x spend on that production speed.

What changed

We switched the first 80% of creative production to AI:

Write 10 prompt variations of the current winner
Generate 30–40 visuals in a single session
Launch 8–12 new ads the same day
Kill losers in 48 hours, iterate winners immediately

The numbers at $50K

Metric	At $5K/mo	At $50K/mo
New creatives / week	2–3	25–40
Time to first test	3–5 days	Same day
Winning ad rate	1 in 8	1 in 4
Creative cost / month	~$2,400	~$29
Blended ROAS	2.1x	2.4x

Spend went up 10x. ROAS didn't collapse — it slightly improved — because creative testing finally kept pace with budget.

The real lesson

More budget without more fresh visuals just burns money faster. AI didn't magically make ads better. It made iteration cheap enough that we could find winners at every spend level.

If your spend is stuck, audit your creative volume before you audit your targeting.

Where is your spend stalled right now — and is creative volume the actual bottleneck? 👇

Hossameldin Ahmed

@heage1·Sep 17

Background removal vs. AI backgrounds — when to use each

Most people default to one or the other. That's leaving conversions on the table. Here's when each actually wins.

Quick decision table

Use case	Use this	Why
Product listing / catalog	Background removal	Clean, consistent, shoppable
Top-of-funnel ads	AI background	Context sells desire
Email product blocks	Background removal	Loads fast, looks native
Lifestyle / brand ads	AI background	Emotion > isolation
Amazon / marketplace	Background removal	Policy + white bg required
Landing page hero	AI background	Story in one frame

When background removal wins

Cutouts are unbeatable for:

Catalog consistency — every SKU on the same white or transparent field
Marketplace compliance — Amazon, eBay, and Google Shopping want isolation
Composable design — drop the product onto any layout without fighting a baked-in scene

Don't overthink it. If the job is "show the product clearly," isolate it.

When AI backgrounds win

AI scenes crush isolated shots when the goal is desire, not identification:

A candle on white = a product
A candle on a rainy-night windowsill = a feeling

That context is why lifestyle AI shots consistently pull 1.5–2x CTR versus cutouts in cold traffic.

The hybrid that actually scales

Shoot or generate a clean isolated product first
Keep that cutout as your source of truth
Generate 5–8 AI environments around it for ads
Keep the isolated version for PDP, email, and marketplaces

Same product. Two jobs. Don't force one treatment to do both.

Which are you overusing right now — cutouts or scenes? Drop it below 👇

Hossameldin Ahmed

@heage1·Sep 17

How to prompt AI for infographic-style visuals that educate and sell

Infographics convert because they teach and sell in one glance. Most AI outputs fail this because people prompt them like photos instead of like layouts.

The infographic prompt formula

"Clean modern infographic on a dark background, [TOPIC] explained in [3–5] numbered steps, simple icons, connecting arrows, large readable labels, generous spacing, no tiny text, professional SaaS marketing design, 4K"

The 5 rules that make it readable

Cap it at 5 steps — more than that becomes noise
Ask for large labels — AI loves microscopic fake text; fight it
Specify layout — "horizontal flowchart" or "vertical numbered list"
Limit colors to 2–3 — brand color + white + one accent
Ban photorealism — add "flat icons, not photorealistic"

Fill-in examples

Process explainer

"Vertical 4-step infographic: Brief → Generate → Test → Scale, blue icons, dark navy background, clean arrows, SaaS style, 4K"

Comparison

"Side-by-side comparison infographic, Stock Photos vs AI Visuals, two columns, green checkmarks and red X marks, bold stats, dark background, 4K"

Stats callout

"Minimal infographic with 3 big stats in circles, short labels under each, lots of negative space, premium dark design, 4K"

Where these outperform photos

Placement	Why infographics win
LinkedIn / B2B feed	Looks like value, not an ad
Carousel ads	One idea per card, high save rate
Landing page mid-scroll	Explains the offer without a wall of text
Email	Scannable proof in one image

Common failure

If the output has fake unreadable text, add:

Negative: "tiny text, lorem ipsum, crowded layout, watermarks, photorealistic people"

Then regenerate. Layout prompts need structure words (steps, columns, arrows) more than style words.

Drop the topic you want visualized and I'll rewrite it as an infographic prompt 👇

Hossameldin Ahmed

@heage1·Sep 16

The minimalist prompt formula for clean, modern brand aesthetics

Minimalism converts. Clean, uncluttered visuals consistently outperform busy compositions for premium and modern brands. Here's the exact formula.

The Minimalist Prompt Formula

"[PRODUCT], centered or off-center on a [NEUTRAL COLOR] background, extremely clean composition, generous negative space, soft single-direction lighting, subtle shadow, muted color palette, no clutter, premium minimalist product photography, 4K"

The 4 non-negotiable elements

1. Neutral color base

Stick to beige, off-white, soft gray, or pale sand. Avoid pure white — it feels clinical, not premium.

2. One light source

Always specify "soft single-direction lighting." Multiple light sources create visual noise that fights minimalism.

3. Generous negative space

At least 40-50% of the frame should be empty. This isn't wasted space — it's what makes the product feel premium.

4. Muted palette restraint

Limit yourself to 2 colors max: the neutral background and one accent from the product itself.

Quick-swap neutral backgrounds

Warm minimalist → "soft sand beige background"
Cool minimalist → "pale stone gray background"
Organic minimalist → "unbleached linen texture background"
Modern minimalist → "soft matte white background with subtle warm undertone"

Why this style performs

Minimalist visuals signal confidence. Brands that need to fill every inch of a frame with detail read as trying too hard to convince you. Clean compositions say "the product speaks for itself."

Common mistake to avoid

Don't confuse minimalist with boring. Add one intentional detail — a subtle shadow shape, a sliver of natural texture, one small prop — to keep it interesting without breaking the calm.

Try this formula on your next product shot and post the result below ✨