9.2 Create Your First AI Image

Section 9: First Steps with AI Reading time: 8 minutes By Thorium-AI Team

This section is a practical, step-by-step walkthrough for creating your first meaningful AI-generated image. We will demystify the process, focusing on actionable techniques over theory, using accessible tools like Bing Image Creator (DALL-E 3) for its ease and Leonardo.Ai for depth. By the end, you will have moved from a blank prompt box to a generated image you can use, understanding the core principles that apply to any image AI.

Phase 1: Foundation – The Anatomy of an Effective Image Prompt

An AI image generator is not a mind reader. It is a pattern completer. Your prompt is a set of instructions telling it which patterns to combine. A strong prompt has four key components:

  1. Subject: The main focal point. Be specific. ("A cat" vs. "a fluffy Maine Coon cat with green eyes sitting upright").
  2. Setting/Environment: Where the subject is. ("on a couch" vs. "on a velvet emerald green Chesterfield couch in a sunlit Victorian library").
  3. Style & Quality Descriptors: How the image should look.
    • Artistic Style: digital art, watercolor painting, oil on canvas, ukiyo-e woodblock print, cyberpunk, Studio Ghibli style, hyperrealistic photography.
    • Quality/Technical Terms: masterpiece, best quality, 4k, hyperdetailed, professional photography, cinematic lighting, sharp focus.
  4. Composition & Parameters (Advanced): Framing, angle, lighting.
    • Composition: close-up portrait, wide-angle shot, low-angle view, Dutch angle, symmetrical composition.
    • Lighting: dramatic rim lighting, soft diffused light, golden hour, neon noir lighting, volumetric fog.

Beginner's Formula: [Subject], in [Environment], [Style], [Quality Descriptor]

Phase 2: Hands-On Tutorial with Bing Image Creator (DALL-E 3)

Tool Choice Rationale: DALL-E 3 excels at interpreting natural language. It's the best tool for beginners to get great results without learning jargon.

Step-by-Step Guide:

  1. Go to bing.com/create and sign in with a Microsoft account.
  2. First Prompt (Simple & Descriptive):

    Enter: A wise old owl with spectacles, reading a large leather-bound book in a cozy, cluttered treehouse library at night. Moonlight streams through a circular window. Digital art.

    Click "Create".

    Observe: You'll get 4 variations. Notice how DALL-E 3 handled the specifics: "wise old owl," "spectacles," "leather-bound book," "treehouse library," "moonlight." It interprets the scene holistically.

  3. Iterate and Refine (The Core Skill):

    Look at the results. Maybe the owl looks too cartoonish, or the library isn't "cozy" enough.

    Refined Prompt: A photorealistic wise old barn owl with tiny round spectacles perched on its beak, intently reading a giant, weathered leather-bound book. It sits in a incredibly cozy, cluttered treehouse library filled with scrolls and glowing orbs, soft moonlight streaming through a round window. Hyperdetailed, magical atmosphere.

    Create again. The new terms (photorealistic, barn owl, weathered, glowing orbs, magical atmosphere) steer the AI toward a different, more detailed result.

  4. Experiment with Style: Take the same core idea and change the style.

    Prompt: [The same owl and library description]... in the style of a Studio Ghibli animated film, soft colors, charming and whimsical.

    Prompt: [The same owl and library description]... as a dark, dramatic oil painting by Caravaggio, chiaroscuro lighting.

Key Takeaway from DALL-E 3: You can speak in sentences. Focus on vivid description. It is forgiving and creative.

Phase 3: Leveling Up with Leonardo.Ai (Control & Precision)

Tool Choice Rationale: Leonardo offers granular control over the model, dimensions, and elements, teaching you the parameters professional generators use.

Step-by-Step Guide:

  1. Go to leonardo.ai, create an account (free), and enter the "AI Image Generation" tab.
  2. Understand the Dashboard:
    • Model Selection: This is crucial. Start with Leonardo Diffusion XL for general purpose or Dreamshaper v7 for realistic people/art.
    • Prompt Box: You will use more structured keywords here, not just sentences.
    • Negative Prompt Box: A powerful feature. Here, you list things you don't want. (e.g., deformed, blurry, bad anatomy, ugly, cartoon, 3d render).
    • Dimensions: Set your aspect ratio (e.g., 1024x768 for landscape).
    • Guidance Scale: How strictly the AI follows your prompt. Start at 7.
    • Tokens: Your generation "currency."
  3. Craft a Structured Prompt:

    Subject/Scene: fantasy warrior queen, intricate armor, glowing sword, standing on a cliff edge

    Style/Descriptors: dynamic pose, dramatic lighting, sunset, epic fantasy book cover art, detailed, by Greg Rutkowski and Artgerm

    Full Prompt: fantasy warrior queen, intricate silver armor, glowing runic sword, standing defiantly on a stormy cliff edge at sunset, dynamic pose, dramatic volumetric lighting, epic fantasy book cover art, highly detailed, digital painting, trending on ArtStation, by Greg Rutkowski and Artgerm

    Negative Prompt: deformed, asymmetric, ugly, disfigured, cartoon, 3d render, plastic, low quality

  4. Generate and Use "Image-to-Image" (Advanced Control):
    • Generate your image. Find one you like but want to adjust.
    • Click on it and select "Image-to-Image."
    • Here's the magic: You can now tweak your prompt while using the generated image as a starting point. Change sunset to aurora borealis or warrior queen to battlemage. Adjust the Creativity Strength slider: high strength changes it a lot, low strength fine-tunes it.

    This allows for controlled iteration, a professional workflow.

Phase 4: Pro Tips & Problem-Solving

Problem: Ugly/Distorted People/Hands.

Solution: Use negative prompts: deformed hands, deformed fingers, mutated hands, extra fingers, bad anatomy. In Leonardo, use a model fine-tuned for people like Dreamshaper v7.

Problem: Image Looks Generic/Like a "Stock Photo".

Solution: Add specific artist names or art movement names to your style (by Alphonse Mucha, in the style of Art Nouveau, French Impressionism). Use unique composition terms (worm's-eye view, dutch angle, symmetrical).

Problem: AI Ignores Part of My Prompt.

Solution: Use weighting. In many generators, you can emphasize words with (brackets:1.5) or COLON:1.2. Example: fantasy warrior queen (with glowing sword:1.3) on a cliff. The number increases its importance.

Ethical Reminder: Be mindful of generating images of real people without consent or in the style of living artists in a way that plagiarizes their work. Use styles as inspiration, not forgery.

Your First Image Challenge:

  1. In Bing Creator, generate: A tranquil Japanese zen garden in the rain, a single stone lantern glowing, cherry blossom petals on wet rocks. Watercolor style, serene mood.
  2. In Leonardo, generate: A sleek, futuristic cyberpunk cat with neon cybernetic implants, sitting on a rainy neon-lit alleyway at night. Blade Runner aesthetic, cinematic.

By completing this guide, you have not just created two images; you have learned the foundational skill of translating mental concepts into a language the AI understands and the iterative process of refinement. This is the core of all AI image generation.

Previous: 9.1 Best Free AI Tools Next: 9.3 Write with ChatGPT
AI Image Generation DALL-E 3 Leonardo.Ai Prompt Engineering Tutorial Digital Art Getting Started