A “negative prompt” refers to a specific type of input that instructs the AI to produce an image with negative or undesirable qualities. Essentially, it asks Midjourney to generate an image that does not align with the desired outcome or intended theme.
This concept is also used on other guide AI models to guide them during the training process to make them more robust and versatile in generating various types of images.
How do I use Midjourney Negative Prompts?
When using negative prompts for image generation on Midjourney, there are several strategies and considerations to keep in mind. We can start a prompt command by typing –no after we have already typed the positive prompts. Here’s a guide on how to effectively use negative prompts:
- Clear Intent and Artistic Direction: Before providing negative prompts, have a clear intent or artistic vision in mind. Understand what you want to achieve with the generated images and how the negative prompts can help you explore new creative avenues or challenge biases.
- Balance with Positive Prompts: Balance negative prompts with positive prompts that specify desired features or concepts. Positive prompts guide the model in the right direction while negative prompts encourage exploration by indicating what to avoid.
- Moderation: Avoid excessive or overly restrictive negative prompts. A good practice is to start with mild constraints and gradually introduce more specific negative prompts as needed. Strive for a balance that allows the model to generate meaningful and coherent images.
- Ethical Considerations: Be mindful of potential ethical concerns. Avoid using negative prompts to generate offensive, harmful, or inappropriate content. Ensure the prompts align with responsible AI use and uphold ethical standards.
- Diverse Negative Prompts: Use a variety of -ve prompts to explore different aspects of the generation space. This can lead to more diverse and intriguing outcomes. For example, instead of only specifying one feature to avoid, combine multiple elements for richer results.
- Experimentation and Iteration: Image generation is an iterative process. Experiment with different negative prompts, observe the results and refine your approach. Iterative refinement helps in achieving the desired artistic output.
- Customization: Customize the negative prompts based on the capabilities and characteristics of the specific model you are using. Different models might respond differently to the same prompt, so understanding the model’s strengths and limitations is essential.
- Context Awareness: Consider the context of the generated images. Negative prompts that are meaningful and relevant to the task at hand will likely produce more coherent and relevant results.
- Feedback Loop: Use the generated images as feedback for improving your prompts. If the model doesn’t produce the desired results, analyze the generated images to understand why and adjust the negative prompts accordingly.
- Documentation: Keep track of the negative prompts used for each image generation. Documentation helps in reproducing successful results and understanding the evolution of the creative process.
By following these guidelines, you can leverage negative prompts effectively to produce unique, creative, and meaningful images through Midjourney. Remember that image generation is an art form, and the thoughtful application of negative prompts can add an exciting dimension to the creative process.
Let’s look at some sample examples that showcase the application of the two negative prompt methods.
Suppose you want to generate an image of a birthday cake without candles. You can use the –no parameter to remove the candles from the generated image:
Original prompt: a birthday cake with candles
Negative prompt: a birthday cake –no candles
Using negative weights
Negative weights offer an alternative approach to specifying unwanted elements in image generation. You can assign negative weights to specific elements in your prompt, making these elements less likely to appear in the output.
To utilize negative weights, you can follow the element’s name with a double colon:: and a negative number (from -0.5 to -1).
For example, if you want to generate a fantasy landscape with fewer mountains and more trees, you can write:
Prompt: /imagine a fantasy landscape using watercolors::0.5 include details like trees::1 mountains::-0.5
How NOT to use it?
While negative prompts can be a valuable tool for encouraging creativity and exploring alternative possibilities in image generation, there are certain situations and contexts where they might not be appropriate or effective. Here are some examples of how not to use negative prompts:
- Overuse: Providing negative prompts too frequently or extensively can hinder the model’s ability to generate meaningful or coherent images. If you excessively restrict the model’s choices, it might struggle to produce anything meaningful, resulting in distorted or unrecognizable outputs.
- Unreasonable Constraints: Using negative prompts that are too extreme or unrealistic can lead to nonsensical or uninterpretable images. For instance, asking the model to “avoid all-natural elements” when generating landscapes might not yield meaningful results and can be counterproductive.
- Bias Reinforcement: While negative prompts can be used to challenge biases in image generation, they can also inadvertently reinforce them. If the negative prompts are based on biased data or assumptions, the model might still generate outputs that perpetuate stereotypes or harmful representations.
- Misunderstood Intent: The model may not always understand the exact intent of a negative prompt, leading to unintended outcomes. For example, if you ask the model to “avoid any human-like features,” it might generate images that appear grotesque or unnatural.
- Arbitrary Constraints: Using negative prompts without a clear purpose or artistic direction can result in random or chaotic images that lack coherence or visual appeal. It’s essential to have a thoughtful rationale behind the use of negative prompts to guide the model effectively.
- Ignoring Positive Prompts: Focusing solely on negative prompts and neglecting positive prompts (instructions about what the image should include) can limit the model’s ability to generate meaningful and balanced images. A combination of positive and negative prompts can lead to more engaging and well-rounded results.
- Ethical Concerns: Negative prompts should not be used to generate offensive, harmful, or inappropriate content. Encouraging the model to avoid certain sensitive topics or offensive imagery is essential to ensuring the responsible use of the technology.
It is important to note that negative prompts may not always result in the desired output. In some cases, Midjourney may still generate images containing unwanted elements despite the inclusion of negative prompts. This limitation highlights the need for experimentation and fine-tuning to achieve optimal results.
Overusing negative prompts or trying to remove too many elements simultaneously can confuse Midjourney and generate unrealistic images. It is essential to use them sparingly and only when necessary.
Below is a list of example negative prompts that you can use for image generation on Midjourney:
- Avoid any human-like features.
- Don’t include any recognizable objects.
- Stay away from symmetrical patterns.
- Exclude bright colors from the image.
- Avoid any sharp or defined edges.
- Don’t use any text or written content.
- Stay away from natural elements like water, trees, or mountains.
- Avoid using any straight lines in the composition.
- Don’t incorporate any animal-like characteristics.
- Exclude any reference to modern technology or gadgets.
- Stay away from the use of shadows or shading.
- Don’t include any references to food or eating.
- Avoid any representations of transportation or vehicles.
- Exclude any reference to historical figures or events.
- Stay away from using any man-made structures or buildings.
- Don’t include any specific cultural or religious symbols.
- Avoid the use of repetition or patterns in the image.
- Exclude any elements that suggest motion or movement.
- Don’t use any elements associated with a particular season (e.g., snow, flowers).
- Avoid any depictions of emotions or facial expressions.
Examples of prompting and results
In the following examples, we showcase the judicious use of negative prompts. In this example, we want to create an image of a gas station in the city limits at night, in rainy weather; our prompt will look like this:
But we don’t want our images to contain nature, humans, or a lens flare. We selected this prompt since the desired image contains a city, its nighttime, and its rain. Midjourney may consider rendering humans and lens flare due to the nature of the prompt. Now, we input the start of a negative prompt: –no and proceed to type the negative prompts separated by a “,” just like a positive prompt, it goes as follows:
Once we are satisfied with the prompt and it follows our artistic direction, we start generating the images, this is the result of the prompt:
You May Also Like:
- MidJourney V5: Take Your Text To Image Prompting Up A Level
- Get FREE Midjourney Fast Hours For Midjourney Fast Mode
- 5 Tips For Midjourney Character Design Prompts:
Remember that the effectiveness of negative prompts depends on the context and the specific model being used. These prompts are intended to spark creativity and encourage the exploration of unconventional ideas during the text-to-image generation process. Feel free to modify and combine these prompts to suit your creative goals and artistic vision.