How to: AI Prompts 101

Welcome to the creative world of AI-powered photo filters! This guide will walk you through the fundamentals of "prompt engineering" so you can design unique and transformative experiences for your photo booth guests.

What Are Booth.Events AI Prompt Filters?

Our AI filters, powered by Google's advanced AI models, allow you to instantly transform guest photos into works of art. The rendered image can be shared through all the regular methods your guests love (SMS, email, Airdrop, etc.).

Important Considerations:

• Processing Time: The time it takes for the AI to generate an image can vary based on the complexity of your prompt and current server load. By default that app will wait until the generation is complete, so that it can immediately show the result to your guests. To prevent lines from forming, we highly recommend using a dedicated capture station with a separate sharing station.

• Beta Feature: This feature is currently in Beta. This means you may see some inconsistencies in the final results. We are continuously working to improve it.
• Group Photos: The AI performs best with individuals and small groups. The larger the group, the more unpredictable the results can become, sometimes leaving people out of the final image. Always test your prompts with a reference photo that has multiple people in the photo, to make sure your prompt works with groups.
• Credits: To use the feature you must utilize credits. When you are on a Pro or Pro+ subscription we get you started with 200 free credits. After that credits can be purchased at any pricing tier but can only be redeemed on Pro, Pro+ & Enterprise plan levels. If you are on a basic plan, you can buy and bank credits and then purchase Pro event credits to be able to use this feature.  For more info on credits see this article.

Understanding Prompts: The Core Concept

A prompt is the set of instructions you give the AI to tell it what kind of image to create. A well-crafted prompt is the key to getting a great result.

A strong prompt is typically built from three main parts: Task, Context, and Reference(s).

• Task: The main action you want the AI to perform. It's the high-level goal.
  • Example: "Turn this photo into..."
• Context: The descriptive details that paint a picture for the AI. This includes the style, setting, lighting, mood, and specific elements you want to see. The more detail, the better.
  • Example: "...a high-contrast black and white photo, with a moody atmosphere."
• Reference: The source image the AI will work from. The photo taken at the booth is the "first" image.

The most critical part of the process is testing and refining your prompt. It is completely normal to try several versions of a prompt to achieve your desired look.

Putting It Into Practice: An Example Walkthrough

Let's break down a prompt to see how these parts work together.

Goal: Turn a regular photo into a classic noir detective movie poster.

Initial Prompt: "Turn this photo into a classic noir detective movie. The subject is under a streetlamp in a rainy alley, wearing a trench coat. High contrast black and white, moody atmosphere, cigarette smoke swirling around."

Let's identify the parts:

• Reference: The image captured by the photo booth.
• Task: "Turn this photo into a classic noir detective movie."
• Context: "The subject is under a streetlamp in a rainy alley, wearing a trench coat. High contrast black and white, moody atmosphere, cigarette smoke swirling around."

Result:

This looks pretty good! But what if we want the image to be more dramatic and focused on the subject's face? This is where iteration comes in.

The Power of Iteration

Let's add more detailed context to our prompt to get a more specific result. The new text is in bold.

Revised Prompt: "Turn this photo into a classic noir detective movie. The subject is under a streetlamp in a rainy alley, wearing a trench coat. High contrast black and white, moody atmosphere, cigar smoke swirling around. The shot is an extreme close-up on their face so you can see all the detail that is lit from the glow of the large dynamic flame of their match they are using to light their cigar that they are puffing."

By providing more specific context about the camera shot and lighting, we get a much more compelling result.

Improved Result:

We could iterate even further by describing the subject's expression, their pose, or their position relative to the camera. Iteration is the key to unlocking the AI's full potential.

Advanced Technique: Using Additional Reference Images

Our Prompt Creator allows you to upload additional reference images to guide the AI. For example, you could provide a reference image for a specific costume style, a background, or an overall artistic look.

How it Works: The AI can process a maximum of three images per request: the original photo booth photo, plus up to 2 of your own. If you add multiple images to a single reference slot, the system will randomly select one of them each time the prompt is run.

When using multiple reference images, you must tell the AI how to use them in your prompt. Remember, the photo of your guest at the booth is the "first" photo, and your reference photos are the "second" and "third" photos.

• Example Prompt: "Turn the subject of the first image into a superhero with a costume inspired by the second image and place them in a futuristic city with a background from the third image."

Best Practices for Success

Follow these tips to write effective prompts and get stunning results.

• Start with Public Prompts: We have created a library of public prompts that are fun and produce great results. Try duplicating one and using it as a starting point: iterate on its style, pose, or subject to fit your needs.
• Use Direct Language: Be clear and confident in your instructions. Avoid ambiguous words like "maybe," "could," or "should." Tell the AI exactly what to do.
• Iterate Constantly: Your first prompt is rarely your last. Start with a base idea, test it, and refine it by adding or changing details until you achieve your vision. Don't be afraid to save versions of prompts that produce good (but not perfect) results in case you need to go back.
• Test with Diverse Subjects: A great prompt works well for everyone. Test your prompt with photos of people with different complexions, hairstyles, and in groups to ensure consistent, high-quality results.
• Be Specific and Detailed: The more specific your descriptions, the better the AI can understand your vision. Think about lighting, color, texture, and mood.
• Specify Image and Art Styles: Tell the AI what style you want. Use terms like: natural unposed street photography, studio photography, fashion photography, shot on a medium format camera, brightly animated like a Pixar movie, photorealistic, watercolor painting, comic book art, etc.
• Define Camera Shots and Angles: Control the composition of your image. Use terms like: worm's-eye view, bird's-eye view, full shot, medium shot, close-up, extreme close-up, etc.
• Use Focus to Your Advantage: Guide the viewer's eye by describing the focus. Use phrases like: "soft focus background," "sharp focus on the subject's eyes," or "create a shallow depth of field" to make your subject pop.
• Use AI to create AI: some users have reported they are using ChatGPT to make their prompts.

A Shot list Cheat sheet

1. Extreme close up
2. Close up
3. Medium close up
4. Medium shot
5. Cowboy shot
6. Medium full shot
7. Full shot