Create Stunning AI Images: A Complete Guide

AI image generation is one of the most exciting features in Family AI Zone. With the right prompts and settings, your family can create artwork, illustrations, custom wallpapers, and more — all from a simple text description.

This guide covers everything from writing your first prompt to understanding every parameter available — size, quality, format, transparency — and how they affect cost and results.

Note: This guide focuses on the OpenAI GPT Image API (official docs). DALL-E 2 and DALL-E 3 are deprecated and will be unsupported after May 12, 2026. Migrate to GPT Image models for best results.

How Image Generation Works

When you submit an image prompt, Family AI Zone sends it to an AI image model (like GPT Image 1.5). The model interprets your text and generates a unique image. The better your prompt, the better the result.

Credits: Each image generation uses credits from your Family AI Zone image quota. The exact number depends on the model, size, and quality you choose. The Owner controls monthly credit limits per member. Credits reset on the 1st of each month.

---

Getting Started: Your First Image

1. From the Dashboard, click Generate Images
2. You'll see the image generation interface with a text input
3. Type a description of what you want to create
4. Select your image model (GPT Image 1.5 recommended for quality)
5. Choose size, quality, and output format (see the Parameters section below)
6. Click Generate

Your image will appear in 5–60 seconds depending on quality and size. You can download it directly to your device.

---

Writing Effective Prompts

The secret to great AI images is a well-structured prompt. Here's the formula:

[Subject] + [Style] + [Lighting] + [Details] + [Mood]

Good vs. Great Prompts

Useful Style Keywords

Art styles:

• watercolor painting — soft, artistic
• digital art — clean, modern
• oil painting — rich textures
• cartoon style — fun, family-friendly
• photorealistic — looks like a photograph
• flat design — minimal, icon-like

Lighting:

• golden hour — warm sunset light
• soft diffused light — gentle, no harsh shadows
• dramatic studio lighting — bold contrasts
• overcast natural light — true-to-life colors

Moods:

• whimsical, cozy, epic, serene, playful, mysterious

---

Prompts for Families

Here are ready-to-use prompts perfect for family use:

For Kids

A magical castle made entirely of candy, with gummy bear guards, rainbow drawbridge, cotton candy clouds, cartoon style, bright colors, friendly and fun

A friendly baby dragon learning to fly, colorful wings, soft watercolor style, white background

For Homework & School Projects

The water cycle diagram — evaporation, condensation, precipitation — illustrated in a clean educational infographic style, labeled

Ancient Rome's Colosseum, historically accurate exterior view, daytime, photorealistic, detailed architecture

For Family Fun

A cozy living room for Christmas Eve, fireplace, decorated tree, stockings, warm light, digital art style

Family game night illustration, board games on a table, warm lamp light, flat design, modern cartoon

---

Choosing the Right Model

Family AI Zone supports multiple image models. Here's how they compare:

Recommendation: Use GPT Image 1.5 Medium for most images. Use GPT Image 1 Mini for children's accounts and simple tasks. Switch to High only for images you plan to print or use prominently.

DALL-E deprecation: DALL-E 2 and DALL-E 3 will stop working after May 12, 2026. Switch to GPT Image models now.

---

Image Parameters Explained

Every image request lets you control several parameters. Each one affects both the result quality and the credits/cost used.

Size

The size determines the pixel dimensions of the generated image. Larger images use more tokens and cost more.

Cost tip: Use 256×256 or 512×512 with DALL-E 2 for the cheapest possible images — great for drafts and children's quick art projects. Each model only supports specific sizes; choosing an incompatible size will show a warning in the UI.

⚠️ Model compatibility: If you select a size unsupported by your current model, the interface will show a warning and block generation. Switch the model or the size to proceed.

---

Quality

Quality controls how many tokens are used to render the image — higher = more detail, more cost.

DALL-E 3 uses standard / hd instead of low/medium/high. DALL-E 2 does not support a quality parameter.

---

Output Format

The format controls the file type returned.

• Use JPEG if you need fast generation and don't need transparency — it's the fastest format.
• Use PNG or WebP if you want a transparent background (great for stickers and logos).
• Compression (0–100%) only applies to JPEG and WebP. Lower = smaller file, slightly less detail.

Format selection is only available for GPT Image models. DALL-E 2/3 always return PNG.

---

Transparent Background

Set background: transparent to generate an image with no background — useful for:

• Stickers and emoji
• Logos and icons
• Overlaying your image onto another design

Requirements:

• Format must be PNG or WebP
• Quality Medium or High gives the best transparency results
• Only supported by GPT Image models

---

Number of Images (n)

You can request 1–4 images in a single generation. This is useful for seeing variations and picking the best one. Credits are multiplied by the number of images requested.

DALL-E 3 only supports n=1. Multiple images are supported by DALL-E 2 and GPT Image models.

---

API Error Messages: What They Mean

If something goes wrong with generation, the app will show a specific message. Here's what to do:

General rule: If you get an unexpected error, try switching to 1024×1024 / Medium / PNG — this configuration works on all models. Then adjust from there.

---

Managing Family Image Credits

As the Family Owner, you control how many images each family member can generate.

Setting Image Quotas

1. Go to Admin Dashboard → Members
2. Click on a member's name
3. Set their Monthly Image Credits (e.g. 20 for children, 50 for adults)
4. Click Save

Monitoring Usage

In Admin Dashboard → Usage, you can see:

• How many images each member generated this month
• Total cost attributable to image generation
• Credit balance remaining per member

What Happens When Credits Run Out?

When a member uses all their image credits:

• The Generate button is disabled
• They see a friendly message: "You've used all your image credits for this month. They'll reset on [date]."
• The Owner can manually add extra credits at any time

---

Tips for the Best Results

Do's ✅

• Be specific about style — watercolor beats artistic
• Describe the scene, not just the subject
• Include the mood/atmosphere
• Mention "family-friendly" or "kid-safe" for content you'll share with children
• Use aspect ratio hints: landscape, portrait, square composition

Don'ts ❌

• Don't request real people by name (GPT Image will usually refuse these)
• Avoid extremely complex scenes with many specific elements — simplify
• Don't repeat words (GPT Image handles this poorly): use synonyms

Using "Negative Prompts"

In the prompt, you can tell the model what to avoid by adding a note at the end:

A peaceful mountain landscape at dawn, no people, no buildings, no text

---

Downloading and Sharing Images

All generated images can be downloaded from the image history:

1. Click Images in the navigation
2. Find the image you want
3. Click the Download button (↓)

Note: AI-generated images are owned by you under the terms of the provider (OpenAI's usage policy applies). You can use them for personal, educational, and most commercial purposes.

---

Editing an Existing Image (Coming Soon)

GPT Image models support uploading an existing image and asking the AI to modify it. This feature is coming soon to Family AI Zone's image generation page.

What You'll Be Able to Do

• Re-style a photo: "Make this look like a watercolor painting"
• Change the background: "Replace the background with a snowy mountain"
• Add elements: "Add a rainbow in the sky"
• Create a scene from a photo: "Turn this pet photo into a cartoon character"

How It Works

1. Upload your image (PNG, JPEG, or WebP; max 3.8 MB, max 2048×2048 pixels)
2. Describe the change you want in the prompt
3. The AI modifies only what you describe, preserving the rest

Inpainting (Selective Editing)

For precise edits, you can upload a mask alongside the image. The mask is a black-and-white image indicating which area to change (white = edit this area, black = leave untouched). This is called inpainting.

Example: upload a room photo and a mask over the sofa → prompt "Replace the sofa with a modern grey sectional."

High-Fidelity Mode

When editing images containing faces or logos, enable High Input Fidelity to ensure the AI preserves those details accurately in the output.

This feature is planned for a future update. When available, you'll see an "Edit Image" tab on the generation page.

---

Cross-Provider Notes

Family AI Zone currently routes image generation through OpenAI's GPT Image API. If your administrator configures a different provider (Stability AI, Flux by Black Forest Labs, Google Imagen), the following differences apply:

Negative prompts workaround for OpenAI: Add what you don't want directly in the prompt:

A peaceful forest, no people, no buildings, no text, no watermarks

Seed parameter: On Stability AI and Flux, providing the same seed with the same prompt reproduces an identical image — useful when you find one you love and want slight variations.

If you're using a non-OpenAI provider and see an error about an unsupported parameter, check the provider's documentation via Admin → Providers and adjust the settings accordingly.

---

Troubleshooting

"Content policy" refused my prompt Your prompt may contain words that trigger safety filters. Try rephrasing — avoid violent, adult, or controversial language. For family accounts, the system applies extra-conservative filtering for child role accounts.

Image looks distorted Try a simpler, more focused prompt. Very long prompts (over 400 words) can confuse the model.

"Size not supported by this model" Each model only supports specific sizes (see the Size table above). Select a compatible size or switch models. DALL-E 2 is the only model supporting 256×256 and 512×512.

"Invalid parameters" or generation failed unexpectedly Try resetting to safe defaults: 1024×1024 / Medium quality / PNG format / n=1. This combination works on all models. Then adjust parameters one at a time to identify which one is causing the issue. Also check Admin → Providers to confirm the model ID and its supported parameters.

Generation is very slow High quality and large sizes take longer — up to 2 minutes for GPT Image High. Switch to Low or Medium quality, or use a smaller size for a faster result.

Image has unwanted artifacts or wrong text GPT Image models sometimes struggle with exact text placement. Add "clean text, sharp lettering" to your prompt, or remove text requests entirely and add text in a separate design tool.

---

Ready to get creative? Head to Image Generation and try your first prompt. Don't be afraid to experiment — generating the perfect image often takes 2–3 tries!