Prompt Engineer 02 - Super Prompts and Prompt Words most used
Mastering the art of prompting can be a game-changer. This comprehensive guide serves as a toolkit to enhance your image prompting skills. Leveraging power words, super prompts, and strategically curated techniques, this handbook provides a foundational blueprint for creating prompts that yield phenomenal results.
Some of these prompts may already be embedded in your AI model, so we recommend trying these out to gauge their effectiveness. As the models continue to improve and evolve, so does the landscape of effective prompting. Hence, this guide is not intended to be a rigid formula, but a dynamic starting point that aligns with the fluidity of this innovative field.
Keep it Simple at First
It's often beneficial to start off with smaller, concise prompts to establish a baseline. This allows you to get a sense of the types of images your AI is currently producing. From there, you can progressively build upon your prompts, adding more detail and specificity to fine-tune the output. When prompt engineering, it's essential to maintain the image seed constant. This approach ensures you're only altering one variable at a time - the prompt - allowing for a clean split test to evaluate what works best and what might require further refinement.
Use your "Super Prompts" to save your images.
Remember, the effectiveness of a prompt doesn't necessarily correlate with its length or complexity. Sometimes, the simplest prompts yield the most powerful results. The key is to test, evaluate, and iteratively refine your prompts.
Moving forward, we'll be zooming in on fine-tuning diffusion models to streamline your prompting experience. If we venture into other models, rest assured, we'll keep you in the loop. If you have any models you'd like us to experiment with, feel free to share. Together, we can navigate the exciting terrain of image prompting, exploring the vast possibilities of AI-driven creativity!
What is a Super Prompt?
A Super Prompt is a power-packed word or phrase that you can incorporate into nearly every image prompt to significantly enhance the final result. These are common terms that can instantly uplift the quality, details, and overall appeal of the generated image.
Examples of Super Prompts include:
"best quality," "masterpiece," "realistic," "detailed," "high quality," "8K," "4K," "ultra-high res," "high detail," "ultra-detailed," "HDR," "RAW photo," "photorealistic," "high resolution," "ultra-realistic," "cinematic," "masterpiece," "DSLR," "gigapixel," "photo-realistic," "hyperrealism," "photorealism," "detailed," "best quality," and "absurdres."
These super prompts may not work with all models, particularly the more modern ones, as they might have these prompts already embedded. Nonetheless, experimenting with one or more of these Super Prompts can potentially yield better outputs if such qualities are not inherently present.
In the following sections, we'll delve deeper into embeddings and negative embeddings, their applications, and how you can utilize them to optimize your prompting process.
Click on Image to Download Cheat Sheet
Prompt Engineer 03 - Embeddings and Negative Embeddings
Embeddings, including negative ones, are powerful tools for influencing the generation of images in AI models. They function as markers or guides that help the model understand the specific characteristics or attributes you desire in your image.
For instance, you can have a 'toy cat' embedding that tells the model you want a representation of a toy cat. Similarly, negative embeddings, like 'bad-artist' or 'bad hands', tell the model to avoid certain qualities, such as poorly drawn figures or anatomically incorrect hands.
You can think of embeddings as adding details to a recipe. Just like you would add instructions like "no garlic" or "extra cheese" to personalize a dish, you can use embeddings to customize your image generation.\
BadDream is good for "dreamshaper style" stuff, while UnrealisticDream is more suited for realistic images, but it's not standalone. You should use it together with BadDream or with other negative words. These can be combined with, Fast Negative Embedding (+ FastNegativeV2)
Using Embeddings to Save Time
One of the key advantages of using embeddings is that they can save a lot of time. Instead of typing out a full description every time you want to generate an image, you can use an embedding. It's a bit like using a shortcut or a code for a specific instruction or set of instructions.
The best part is that once you have created an embedding, you can save it and use it again whenever you need it. So, if you find yourself often generating images with specific characteristics, creating and using embeddings can make your work much more efficient.
Popular Embeddings Below:
For your convenience, we will provide a list of commonly used prompts and negative prompts below. This list can serve as a useful reference, helping you to get started with using embeddings to generate fantastic images!
These are the most commonly used prompts that does magic to your image generations that I call "Super Prompts."
Click on Image to Download Cheat Sheet
1. Quality and Resolution
HDR RAW photo
Photorealistic High Resolution (Photography)
Ultra-realistic Cinematic Masterpiece
2. Specific Details and Features
Cold skin pores
Supercharger on car
45-degree angle shots
Sleek and aerodynamic body (car)
Large and sturdy wheels (car)
Dragon scaly armor
3. Artistic Techniques & Technology
Radiosity 3D Render
4. Genre and Style
George Miller's Mad Max style
5. Settings and Environment
Ancient fallen kingdom
Office with wall of monitors and computers
Post-nuclear fallout setting
Empty desert highway
Tunnel in Hell Creepy Terrifying
6. Lighting and Effects
Teal and orange
7. Artists and Platforms
8. Colors and Tones
Natural skin texture
Black color (car)
Spots of dust and dirt (car)
Rugged and gritty (car)
9. Specific Views and Angles
Reflections in water
Half body portrait
Full body photo
View from back
Medium upper body shot
Looking at the camera
10. Cameras and Photography Techniques
High speed camera
Black and white
Film noir style
Soft glow effect
Soft pastel colors
11. Negative Attributes
Blurry Bad anatomy Low quality Lowres Overexposure Sketch Artifacts Washed-out Bad proportions Extra limbs Disfigured Deformed Poorly drawn hands, face Extra fingers, arms, legs Long neck Gross proportions Missing arms, legs Mutated hands Fused fingers
12. Negative Format and Technical Issues
Low quality Lowres Jpeg artifacts Artifacts Glitch Error Out of focus Amateur Poorly lit Overexposed Underexposed Cropped Out of frame Semi-realistic Cartoon Drawing Digital art Anime Manga Painting Doll Fake 3D modeling Greyscale Monochrome
13. Negative Anatomy and Proportions
Poorly drawn hands Poorly drawn face Poorly drawn feet Poorly drawn eyes Deformed iris Deformed pupils Deformed face Bad teeth Bad hands Bad fingers Bad eyes Long body Bad anatomy Gross proportions Extra fingers Too many fingers Extra limbs Extra arms Extra legs Missing arms Missing legs Fused fingers Long neck Mutated hands and fingers Conjoined Missing limbs Mutation Deformed Dehydrated Disfigured Mutated Bad proportions Malformed limbs
14. Negative Subject Matter and Miscellaneous
Asian, Japanese, Korean, Chinese Child Strabismus Plump Fat Muscular female Man with scruffy beard, balding, chubby face, wide chin, squinting, glasses Man with dirty grey hoodie and sweatpants Twisted Distorted Logo Watermark Signature Copyright Title Text Trademark Username Cloned Ugly Gross Morbid Mutilated Incoherent
15. Negative Lighting and Effects
Noisy Pixelated Fish-eye lens distortion Distorted perspective Uneven lighting Harsh flash Red-eye effect Motion blur Over saturation Unnatural colors Visible tripod Reflection of the photographer Visible dust or smudges on the lens Overly complex and confusing composition Awkward pose Obvious photo manipulation Beautiful form of chaos
21charturnerv Asian-Less-Neg CyberRealistic_Negative-neg EasyNegativeV2 HyperStylizeV6 bad-artists-anime bad-artist bad-hands-5 bad-image-v2-39000 bad-picture-chill-75v bad-pictures bad_prompt bad-prompt_version2 badhandv4 charturnerv2 easynegative nartfixer negative_hand-neg nfixer ng_deepnegative_v1_75t nrealfixer pureerosface_v1 rmadanegative402_sd15-neg ulzzang-6500-v1.1 ulzzang-6500 verybadimagenegative_v1
To wrap it up, the power and beauty of prompting is limited only by your imagination and creativity. Whether you're diving into the basics of prompting, enhancing your skills with super prompts, or venturing into the advanced territory of embeddings and negative embeddings, this guide is designed to take you through every step of the journey.
Remember, there's no "one size fits all" in this rapidly evolving landscape. With the continual improvements and evolution in AI models, we encourage you to experiment with prompts and negative prompts, to make each image better and more efficient. Start small, gauge the results, and gradually build on that foundation. Retain the same image seed when prompt engineering, to effectively split test what works best for you.
And finally, don't forget the importance of time-efficiency. Make full use of embeddings as preset groups, which you can reuse over time, making the process of prompt engineering faster and easier.
This is just the beginning. As you traverse through the world of AI and image prompting, keep experimenting, learning, and above all, have fun. After all, the journey is just as important as the destination.
Until next time, keep prompting!
Help Support this Page
Please click on the image below to download this helpful Cheat Sheet. By sharing this page or the Cheat Sheet on Twitter @AndyHTu, you can contribute to the expansion of this resource and ensure we can continue providing you with valuable training material for future AI Photography Workflows, including the use of Stable Diffusion and other cutting-edge tools. Your support is greatly appreciated and instrumental in our growth.
Visit Civitai Community to download fine tuned models to play with.
Reference: Github Automatic1111 User Guide.
Stay up to date with what's happening with Stability AI and Stable Diffusion.