What customers ask AI for:

• “BIRTHDAY GIRL” → Get “BIRTDHAY GRIL”
• “CLASS OF 2026” → Get “CLARS OF 20E6”
• “WORLD’S BEST DAD” → Get “WQRLD’S BSET DAD”
• “TEAM WILSON” → Get “TEAM WILSNO”
• “BRIDE TRIBE” → Get “BRIED TRBIE”

And that’s when AI bothers to include text at all. Sometimes it just… doesn’t. Or puts in decorative squiggles that look like letters but aren’t.

1. The AI sees your prompt: “Create a design with ‘HAPPY BIRTHDAY’ text”
2. It recognizes a pattern: “Birthday designs often have text-like shapes”
3. It generates visual noise: Shapes that look like letters based on training data
4. It doesn’t verify spelling: There’s no spell-checker. It’s just making shapes.

Diffusion models work pixel-by-pixel:

• They start with visual noise and gradually refine it
• They predict what pixels should be where based on training data
• They understand “text shapes exist here” but not “these letters spell a word”
• There’s no text rendering engine involved

Language models vs Image models:

• ChatGPT (text) understands language, grammar, spelling
• DALL-E (images) understands shapes, colors, composition
• They’re different systems with different capabilities
• Even though they’re integrated, DALL-E doesn’t inherit ChatGPT’s language skills

Current state (2026):

• Very short text (2-3 letters) sometimes works
• Single words occasionally work
• Phrases and sentences almost never work
• Complex layouts with multiple text elements? Forget it.

Even when text is technically spelled correctly, it often has weird spacing, inconsistent sizing, or bizarre angles. You can’t control font, weight, or positioning reliably.

• You need specific text: Names, dates, phrases that must be exactly right
• You need multiple text elements: Title + subtitle + tagline, etc.
• You need text formatting control: Specific fonts, sizes, colors, alignment
• Spelling matters: Anything professional, official, or gifted
• You’ve already tried 3+ times: It’s not getting better. Save yourself.

• Text is decorative/illegible anyway: Grunge/distressed styles where “readability” isn’t the point
• It’s background texture: Newspaper print effect, script background patterns
• Foreign/fantasy languages: When authenticity isn’t verifiable
• Very short words: “LOVE,” “BEER,” “OK” — sometimes works
• The design is clearly AI art: And text mistakes are part of the aesthetic

❌ Bad prompt:

“Create a birthday design that says ‘HAPPY BIRTHDAY SARAH’ with balloons and confetti”

✅ Good prompt:

“Create a birthday design with balloons and confetti, bold graphic style, no text, leave space at the top for text to be added later, on pure white background”

Key phrase: “no text” or “without any words” — be explicit about this.

• Free version works great — No need for Pro for basic text
• Intuitive interface — If you can use PowerPoint, you can use Canva
• Professional fonts included — Hundreds of options
• Text effects built in — Outlines, shadows, curves, arches
• Exports high resolution — Set to PNG, 300 DPI for printing
• Transparent background support — Essential for printing

1. Create custom size canvas:

• Go to Canva.com
• Click “Create a design” → “Custom size”
• Set to your desired final print size (e.g., 12″ x 12″ for large chest print)
• Make sure it’s set to at least 300 DPI

2. Upload your AI image:

• Click “Uploads” in left sidebar
• Upload your AI-generated image
• Drag it onto the canvas
• Resize to fit (leave room for text if needed)

3. Add text:

• Click “Text” in left sidebar
• Choose “Add a heading” (larger) or “Add a subheading” (smaller)
• Type your actual text
• Position it where you want it

4. Format text properly:

• Select the text box
• Choose font from dropdown (search for specific styles)
• Adjust size using slider or typing exact pt size
• Set color to match or contrast with your design
• Add effects if needed (Effects → Outline, Shadow, Lift, etc.)

5. Download for printing:

• Click “Share” → “Download”
• File type: PNG (with transparent background if applicable)
• Quality: Check “Use print quality (300 DPI)” if available
• Download

If your AI design is:

Bold, graphic, comic-book style →

Use: Impact, Bebas Neue, Oswald, Anton, heavy sans-serifs

Retro, vintage, 70s aesthetic →

Use: Cooper Black, Groovy, Funkydori, rounded bold fonts

Elegant, watercolor, soft →

Use: Script fonts (Pacifico, Great Vibes), serif fonts (Playfair Display)

Modern, minimalist, geometric →

Use: Montserrat, Raleway, Futura, clean sans-serifs

Grunge, distressed, edgy →

Use: Distressed fonts, hand-drawn styles, rough textures

Cute, playful, kids’ style →

Use: Rounded fonts (Fredoka One, Baloo), bubbly styles

• Bebas Neue (title) + Montserrat (subtitle) — Modern, professional
• Impact (main text) + Arial (supporting text) — Bold, readable, classic
• Oswald (title) + Open Sans (body) — Clean, versatile
• Anton (big statement) + Roboto (details) — Strong contrast
• Pacifico (script) + Lato (sans-serif) — Elegant with readability

• Overly decorative fonts — Hard to read, especially at distance
• Too many fonts — Stick to 2 maximum (title + subtitle)
• Thin, delicate fonts — Disappear on busy backgrounds
• Comic Sans — Just… don’t. Unless it’s ironic.
• All caps in script fonts — Designed for mixed case, looks weird all caps. Yes, we know you’re yelling. Script fonts still don’t work that way.
• Stretched or squashed text — Distorts letterforms, looks unprofessional

1. Outlines/Strokes (Most Important)

Adds a contrasting border around letters. Essential for text on busy backgrounds. In Canva: Effects → Outline → adjust thickness and color. White text? Add dark outline. Dark text? Add white outline.

2. Drop Shadows

Creates depth, makes text “float” above the background. In Canva: Effects → Shadow → adjust blur and offset. Subtle works best.

3. Background Shapes

Place a semi-transparent rectangle, circle, or banner shape behind text. Guarantees readability. Elements → Shapes → choose shape → place behind text → adjust transparency.

4. Lift/3D Effects

Adds dimension. Works well for bold, playful designs. In Canva: Effects → Lift → choose color for the “depth” layer.

Outline: When background is busy/multicolor and text needs maximum visibility

Shadow: When you want depth but background isn’t too busy

Background shape: When outline alone isn’t enough, or for a “badge” aesthetic

Lift/3D: For fun, casual designs where you want boldness (kids’ shirts, event tees)

• Light background → Dark text (black, navy, dark gray)
• Dark background → Light text (white, cream, light gray)
• Busy background → Text with outline in contrasting color
• Test readability: Zoom out. If you squint and can’t read it, change it.

Above the graphic:

Classic, formal. Works well for: team names, event titles, organization names. Example: “LINCOLN HIGH SCHOOL” above school mascot.

Below the graphic:

Grounding, explanatory. Works well for: taglines, dates, locations. Example: Mountain graphic with “EST. 2020” below.

Arched above:

Traditional, badge-style. Works well for: logos, official-looking designs, circular graphics. In Canva: select text → Effects → Curve.

Integrated/overlapping:

Modern, dynamic. Text partially overlaps the graphic. Requires good contrast/outline. Works well for: bold statements, modern designs.

Banner across middle:

Ribbon/banner effect through the center. Works well for: names, featured text, special occasions.

• Main text (title): Should be readable from 6-10 feet away
• Subtitle/supporting text: Can be smaller but must still be legible up close
• Minimum text size: 0.25 inches tall (about 18pt at print size) for readability
• Don’t make text too small: If someone over 50 can’t read it, it’s too small
• Balance with graphic: Text shouldn’t overpower the AI image or get lost

• Leave breathing room: Don’t cram text right against graphic edges
• Line spacing (leading): Multi-line text needs space between lines for readability
• Letter spacing (kerning): All-caps text often benefits from slightly increased spacing
• Margins: Keep text away from edges of the print area (0.5-1 inch minimum)

GIMP (Free Photoshop alternative)

• Pros: Powerful, professional features, true layer control
• Cons: Steeper learning curve, less intuitive interface
• Best for: People comfortable with advanced editing

Photopea (Browser-based Photoshop clone)

• Pros: No download needed, familiar Photoshop-style interface
• Cons: Ads in free version, requires internet
• Best for: Quick edits without installing software

Pixlr (Simplified online editor)

• Pros: Easier than GIMP, still capable
• Cons: Limited fonts in free version
• Best for: Basic text addition

Adobe Photoshop

• Cost: $20-55/month depending on plan
• Best for: Professional designers, people who need industry-standard tools
• Overkill for: Just adding text to AI images for t-shirts

Affinity Designer/Photo

• Cost: One-time purchase ~$70
• Best for: People who want professional tools without subscription
• Learning curve: Moderate

❌ Mistake #1: Text too small to read

Fix: Text should be minimum 0.25″ tall. Test by viewing at actual print size.

❌ Mistake #2: No contrast with background

Fix: Add an outline/stroke in contrasting color, or place text on a background shape.

❌ Mistake #3: Text at weird angles

Fix: Horizontal or slightly arched text reads best. Avoid dramatic angles.

❌ Mistake #4: Stretched/squashed text

Fix: Hold Shift when resizing to maintain proportions, or use font size slider.

❌ Mistake #5: Too many fonts

Fix: Stick to 2 fonts maximum: one for main text, one for supporting text.

❌ Mistake #6: Low resolution export

Fix: Always export at 300 DPI for printing. Check Canva download settings.

The Smart Workflow:

1. Generate AI image WITHOUT text
2. Upload to Canva (or similar tool)
3. Add text properly with real fonts and formatting control
4. Export at 300 DPI
5. Send to printing