midjourney-tutorial
Using Your Own Images in Midjourney: Style Reference, Describe, and Image Prompts
Midjourney AI Team · July 23, 2026 · 7 min read
Keywords: midjourney image prompt, style reference midjourney, midjourney describe command
Published: July 23, 2026 Author: Midjourney AI Team
Who This Guide Is For
Integrating your own assets into generative workflows is no longer optional for serious creators; it is a requirement for brand consistency and iterative design. This guide is written for visual designers, brand managers, and digital artists who need to move beyond random generation. If you are trying to maintain a specific aesthetic across multiple campaigns or need to iterate on a specific character without losing their identity, standard text prompting will not suffice. You need control over composition, color palette, and structural integrity.
Using image inputs in Midjourney allows you to anchor the AI's imagination to reality. Whether you are uploading a mood board, a sketch, or a previous generation, understanding the hierarchy of image commands prevents wasted iterations. We will cover the technical distinctions between Image Prompts, Style References, and Character References, ensuring you know exactly which parameter to deploy for your specific goal. You can test these workflows directly in a managed environment by visiting the studio.
Quick Takeaways
Mastering Basic Image Prompts
Before diving into specialized reference parameters, you must understand the foundational behavior of image prompts. When you attach an image URL to your text prompt without any specific flags, Midjourney treats the image as a strong compositional guide. The AI analyzes the colors, shapes, and general layout of the uploaded file and attempts to replicate that structure in the new generation.
This method is highly effective for mood matching. If you have a photograph with specific lighting conditions—say, a golden hour portrait with heavy shadows—you can use that image to dictate the lighting of a completely different subject. However, basic image prompts can be unpredictable. The AI might cling too tightly to the original content, resulting in variations that look too similar to the source rather than inspired by it.
To manage this, you can adjust the image weight. While the specific --iw parameter has evolved across versions, the principle remains: you are balancing the influence of the image against your text description. In version 7 (--v 7), the model is more sensitive to text, so your image prompt acts more as a stylistic anchor than a rigid template. For best results, combine your image URL with descriptive text that clarifies what should change. For example, uploading a sketch of a car and prompting "cyberpunk vehicle, neon lights, rainy street" tells the AI to keep the shape but change the texture and environment.
Style Reference for Brand Consistency
The Style Reference parameter (--sref) is arguably the most powerful tool for maintaining visual identity. Unlike a basic image prompt, which influences composition and content, --sref isolates the aesthetic qualities of an image. It extracts color palettes, texture patterns, and rendering styles while ignoring the subject matter.
If you are building a brand deck, you can upload a key visual asset as your style reference. When you generate new images using --sref [image_url], the new outputs will adopt the look and feel of that asset without copying its content. This is crucial for marketing teams that need diverse imagery that still feels like it belongs to the same campaign.
For maximum control, pair --sref with --style raw. The raw mode reduces Midjourney's default beautification filters, allowing the style reference to dictate the aesthetic more accurately without the model imposing its own artistic bias. You can also adjust the style weight using --sw. A higher weight forces the style more aggressively, while a lower weight allows for more subtle integration. When working in MidassAI Studio, you can manage these references easily without needing to host images externally, streamlining the iteration process.
Character and Omni Reference Techniques
Maintaining character consistency across different scenes is a common hurdle in generative art. The Character Reference parameter (--cref) addresses this by focusing on facial features and clothing details. When you provide a reference image of a character, Midjourney attempts to map those features onto new generations.
To use this effectively, ensure your reference image has a clear, unobstructed view of the character's face. Side profiles or heavily obscured faces often confuse the model. You can control the strength of the character retention with --cw. Setting --cw 0 focuses only on the face, allowing you to change outfits and hair easily. Setting --cw 100 attempts to copy the entire look, including clothing and hair style.
Recently, the Omni Reference (--oref) has emerged as a comprehensive solution. This parameter combines the capabilities of style, character, and composition into a single flag. It is designed to understand the image holistically. While --sref and --cref offer granular control, --oref is useful when you want the AI to interpret the image's intent broadly. However, for professional workflows where specific elements need to remain static while others change, the dedicated --cref and --sref parameters often yield more predictable results.
The Describe Command for Reverse Engineering
Sometimes you have an image but lack the vocabulary to describe why it works. The /describe command is a reverse-engineering tool that analyzes an uploaded image and suggests four potential text prompts. This is invaluable for learning how to construct prompts that yield similar results.
When you upload an image to /describe, Midjourney breaks down the visual elements into keywords, artistic styles, and camera parameters. You might discover that the specific look you admire is due to a "35mm lens" or "volumetric lighting" tag you hadn't considered. Use these suggested prompts as a starting point, then modify them to suit your needs. This command bridges the gap between visual intuition and textual precision, helping you build a library of effective prompt structures based on existing assets you admire.
Generating Video From Static Images
The ecosystem around generative AI is expanding beyond static frames. While Midjourney focuses on high-fidelity imagery, the broader MidassAI Studio environment allows you to take those static generations and animate them. Once you have perfected your image using the reference techniques above, you can transition those assets into video workflows.
Static images serve as excellent starting frames for video generation. By using a consistent character reference (--cref) to create a sequence of images, you can then feed those into video models to create coherent motion. This workflow ensures that your video content maintains the same visual fidelity and character consistency as your static marketing materials. It is a multi-step process: generate the keyframes in Midjourney, ensure consistency with style references, and then animate within the studio's video tools.
Tips for Best Results
Achieving professional output requires attention to detail beyond just the parameters. First, consider the aspect ratio (--ar). If your style reference is a vertical portrait but you are generating a horizontal banner, the AI may struggle to map the style correctly. Try to match the aspect ratio of your reference image to your generation target where possible.
Lighting consistency is another critical factor. If your style reference uses flat, studio lighting, do not expect dramatic shadows in your output unless you specify them in the text prompt. The AI tries to harmonize the text and image inputs; conflicting instructions lead to muddy results.
Finally, version control matters. Always specify the model version you are working with, such as --v 7. Newer models handle references differently than older ones. A style reference that worked perfectly in version 5.2 might behave differently in version 6 or 7. Document your successful parameter combinations. Keep a log of which --sw or --cw values worked for specific projects. This institutional knowledge saves time on future campaigns.
Next Steps in Your Workflow
Mastering image inputs transforms Midjourney from a novelty into a production tool. By leveraging --sref for aesthetics and --cref for identity, you gain the consistency required for commercial work. The /describe command further accelerates your learning curve, turning visual analysis into actionable text.
To streamline this process and manage your assets effectively, consider using a dedicated platform. You can experiment with these parameters and manage your generations in a unified interface.