Midjourney is taking the world of AI art generation by storm. With just a text prompt it can create stunning and creative images. One of the most powerful features, however, is the ability to use reference photos to influence the output. This opens up endless possibilities for generating custom AI artworks. In this comprehensive guide, we explore the ins and outs of using reference images in Midjourney to create visual masterpieces.
What are reference photos during the trip?
Reference photos act as visual examples that guide the Midjourney AI in generating an image. By uploading relevant photos, you provide concrete examples of things like colors, lighting, poses, styles, compositions, and more that you want to include in the final output. Reference photos help determine the AI’s artistic direction.
Why use reference photos?
Here are some of the key benefits of using reference images in Midjourney:
- Provides visual inspiration for the AI – photos act as a creative springboard.
- Gives control over colors, mood, lighting, poses, etc. based on sample photos.
- Ensures that important visual elements are included in the final artwork.
- Ensures consistency between multiple generations using the same references.
- Makes it easier to communicate abstract concepts visually.
- Adds visual dimensionality that can be difficult to convey in text alone.
In general, reference photos allow you to dial in exactly the visual aspects you want in your AI art. The AI brings it all together in new and original ways.
How to use reference photos mid-trip?
Step 1: Find good reference photos
The first step is to gather a strong library of reference photos tailored to your needs. Here are a few tips:
- Search online image platforms like Unsplash or Pexels for relevant, high-quality photos.
- Look for images that visually represent the final output you want to create.
- Create a collection of 5-10 photos to highlight different angles, perspectives, styles, etc.
- Save or download the photos, ready to upload.
Aim for 5-10 great photos that really visualize the most important aspects of the final artwork you have in mind.
Step 2: Upload reference photos to Midjourney
Once you have your reference photos ready, it’s time to upload them:
- Launch Midjourney and open the Discord chat box to access the AI bot.
- Drag your photos directly into the Discord chat box to upload them.
- You can also copy and paste the file paths or image URLs into the text prompt box.
- Midjourney processes and uploads each image as you go for use in art generation.
- You can upload up to five reference photos at a time for one generation.
Make sure your photos have been successfully uploaded before continuing with creating clues.
Step 3: Write prompts with references
Now that your credentials have been uploaded, it’s time to write your generation prompt. Prompts should:
- First clearly describe the desired final output at a high level. For example: ‘An oil portrait of a girl with a pearl earring’.
- Mention individual reference photos and what you would like to borrow from each photo. Such as “The girl has the facial features and expression from reference 1”.
- Describe changes or new elements that are not included in the references. Such as ‘She’s wearing a lace dress’ or ‘In an ornate gold picture frame’.
- Provide clear and concise language, but with enough detail for the AI to interpret.
- Use the references to communicate styles, lighting, moods, composition, etc.
Taking the time to carefully reference each uploaded image in your prompt will ensure that the AI understands how to weave them into an original creation.
Step 4: Adjust image weights
An important parameter when using references is the image weight. This determines how closely the AI adheres to the visual elements in a reference when generating art.
You can set custom weights when calling images in your prompt:
- By default, Midjourney gives reference photos a weight of 1.
- If you want the output to closely match a photo, set the weight to 1.5 or 2.
- For looser inspiration, set the weight lower, for example 0.5.
- Experiment with different weights to achieve the ideal image mix.
Experiment with different image weights in your prompts to find the right mix between sticking to references and injecting creativity. There are no right answers here – follow your creative instincts!
Step 5: Remix and Reroll
Once your first image has been generated using references, you can further customize and iterate using Midjourney’s ‘remix’ feature.
Remixing allows you to take an existing image and generate stylized variations. Some ways to remix with references:
- Press Enter to have Midjourney generate an image based on your prompt and references.
- If necessary, activate remix mode and reroll for variations on your original output.
- Try adjusting parameters such as aspect ratio, sampling method, etc.
- Remix and repeat until you achieve the desired result.
Remixing creates an iterative creative process in which you can adjust the balance between reference influence and AI artistry until you achieve your ideal vision.
Step 6: Combine multiple references
Where Midjourney really shines is in combining multiple reference photos with text prompts to weave disparate concepts into cohesive works of art.
You can combine references from things like:
- Once you have a good basic output, you can create stylistic variations with different cues and photo combinations.
- Change the mood, perspective, color schemes, etc.
- Generate multiple images to build a cohesive AI art series.
The AI does the heavy lifting of putting these visual elements together into a unified artistic creation. The results are often new and unexpected.
Some tips when combining multiple visual references:
- Consciously manage references so they have a contextual connection. Don’t choose random photos.
- Use references that reflect similar perspectives, depth, and framing for greater coherence.
- Write prompts to bridge the gap between seemingly disconnected visual elements.
Through thoughtful selection and prompting, mixed references produce original and imaginative AI artworks.
Quick inspiration with /Imagine
Midjourney’s /imagine command provides another way to include references in new clues. For example:
/imagine this photo as a renaissance oil painting
The AI generates a new interpretation of a reference photo in the specified style. You can draw inspiration from the output to guide new prompts and creations.
Experiment with different references
Ultimately, using references effectively requires experimentation. Each combination of image, prompt, and reference produces different results.
Discover what works best through trial and error. Make small iterative changes and compare the results. Over time, you will develop intuition for which references work well together.
Also take risks with unexpected combinations of references and clues. Happy accidents often lead to the most unique AI creations.
The world of AI art generation is yours for the taking once you master credentialing. Let your imagination run wild!
References provide invaluable context and direction for Midjourney to turn your imaginative ideas into reality. With the right photos and prompts, you can create deep, meaningful art that conveys exact moods, styles, and concepts.
Follow the tips in this guide to choose eye-catching references, write detailed prompts, adjust image weights, and remix generative parameters. Mastering references unlocks Midjourney’s full potential for creating professional AI artwork infused with your personal vision.
So gather some inspiring references, type out the clues and let Midjourney work its algorithmic magic! Bring your creative dreams to life with AI and reference photos. The possibilities are endless.