How to Generate AI Images: From Concept to Output in 5 Steps
In this beginner-friendly guide I will walk you through creating professional AI images using tools like Midjourney, Runway, Freepik, and ChatGPT. Whether you're a complete beginner or looking to improve your AI image generation skills, this 5-step workflow will help you create images that actually look good.
Workflow Tools
Quick links to the tools mentioned in this AI image generation guide.
1
Choose Your Tool
- If you want the best results, pay for Midjourney. It delivers the highest quality AI generated images available today (not a free option though).
- For beginners learning how to create AI images for free, I found Runway to be an amazing choice. While most people know it for AI video, their image generation tool offers 300+ credits (one image costs 8 credits).
- Runway vs MidJouruney difference at the end
2
Upload References and Define Your Style
- Both Runway and Midjourney let you upload reference images. These show what you want in your final image. Upload images of a knight + meadow = knight in a meadow output.
- Styles control the vibe and help keep consistency across images.
3
Looking for Inspiration and References
- I find it super helpful to look for inspiration and references when generating AI imgaes. I use Freepik to find reference materials.
- It has a big library of art, photographs and AI images.
- Freepik lets you filter by uploading an image, I find it super useful when looking for similar images or styles to the one I like.
4
Write Your Prompts
- Most people treat AI image prompting like Google search, typing basic keywords. It works, but you definitely won't get amazing results. Below is the template I use. You can just fill ill in: what you want + mood + key details + camera angle + colors.
- I want to generate a [describe what you want to generate] with a mood that feels [mood or atmosphere]. The image should include [key details or elements], and the camera angle should be [select camera angle]. I'd like the color palette to feature [preferred colors].
- After writing the initial prompt, I then run it through ChatGPT to improve it. Just swap the original prompt and AI image tool you’re using.
- Refine the following prompt: [Insert your prompt here]. Ensure it adheres to the guidelines below: Your task is to refine and enhance image generation prompts for Runway Gen-4 AI. Follow these guidelines strictly: Use descriptive, positive, and natural language. Avoid vague or conversational phrasing. Include details about subject, scene, composition, lighting, color, style, focus, angle, text, and mood. Make the prompt visually rich, emphasizing intricate details and stylistic elements. Avoid using negative phrasing. Instead of describing what shouldn’t appear, focus on what should. Keep the final prompt concise, under 1,000 characters, while ensuring it is detailed enough to guide the model effectively. For example: Original prompt: 'a woman in a veil in a dreamy setting' Improved prompt: 'elegant portrait of a woman draped in a flowing sheer tulle veil against a warm glowing sunset sky. Close-up composition, ethereal atmosphere with soft, dreamy lighting. The veil flows weightlessly, creating dynamic movement in the scene. Subtle pastel and vibrant tones blend together, evoking a baroque painting style. Glowing peach and golden accents add a whimsical touch.
- Runway struggles with camera angles. I upload the final image from Runway to Flair.AI when I need specific angles, it handles them well (ChatGPT works too)
5
Runway AI vs MidJourney
- As you can see, MidJourney can deliver much more realisitic result. The lowest membership for MidJourney is 10$ a month. You can decide if it's worh it for you
- Personally, if you wanna actually learn AI Art generation, 10$ for the best tool out there is worth it (not affiliated with MidJourney).
- At last: Keep practicing. Most of the time it's not the AI that's bad, it's that we don't know how to prompt it well enough.