My capstone project is an interactive website centered around a creepypasta(the horror-related legend that has been shared around the Internet) surrounding a fictional hotel. Within the confines of this immersive web platform, users start on a journey of discovery of the eerie secret concealed within the hotel. In the process of creating this website, I have employed multiple AIs to create text, images, code, and even videos in the future. Without the assistance of AI tools, it is impossible for me to solely produce this much content in a given amount of time.
Adobe Firefly is a very powerful image generative AI tool and it is commercially safe. it claims that they are training our initial commercial Firefly model on licensed content, such as Adobe Stock, and public domain content where copyright has expired. It has four main functions: text-to-images, text effects, generative fill, and generative color. My most common practice is using text to images to generate images of hotel room, dining, rooftops....The reason for using AI is not only to create photo realistic images but more importantly photo-realistic "unreal" and "bizarre" images, which fit the eerie, fake vibe of the hotel website. For example, I can input the prompt "Create a photo realistic luxurious hotel rooftop with the view of hotel and forest". The output image of hotel sitting in a forest New York is pretty interesting.
After selecting the images I can go to generative fill to add and subtract unwanting parts of the images, and ask AI to generate a new part that fits seamlessly to the original image. For example, if I dislike the furniture here, I can select them and ask AI to a "modern luxurious sofa and antique wood table". Ai will then replace the old sofa with a new one generated according to my prompt.
I've been relying on Chat GPT nearly every day for this project. Primarily, I use it to generate descriptive text that mimics the style of modern hotel website marketing slogans, as I have limited knowledge in that area. I've included a few examples of me asking Chat GPT to craft a slogan for the hotel banner, as well as marketing text for rooms, dining, and other amenities. Among the suggestions, I particularly favor the slogan "Every Moment is Unforgettable." It's concise, straightforward yet memorable, and intriguingly, it seems to foreshadow the unforgettable secrets that unfold within the confines of the hotel.
I used to use chatgpt a lot until recently I encountered with co pilot. It is free for student and it is super powerful as it is learning from all github repository. It is also very convience because it can be employed directly in editor such as Visual Studio Code. Copilot can provide real time ai suggestions. It can predict what you want to write from your previous code and all you need to do is press tab to accept that suggestion. The copilot chat act like chatgpt that can analyze selected code in the file and answer your question. With copilot, I can write much more faster code as it can correct my code, write my code and even teach me how to write code.
When it comes to AI assistance, my feelings are quite mixed. On one hand, I recognize that without AI's aid, undertaking this project solo would be a nearly impossible task. The prospect of manually photographing existing hotels and then photoshopping them seems daunting. Creating videos is nearly impossible; I might have to 3D model the hotel and generate a video from scratch. Not to mention, without tools like Copilot and ChatGPT, I'd likely spend hours learning new features, unsure if they'd align with my vision. Debugging, too, could easily devour countless hours of my time. AI tools allow me to rapidly build up my prototype and test my ideas.
At the same time, I do acknowledge that AI models are trained on vast amounts of online data, gathering information, articles, and artists' works without their consent. As a visual artist, I deliberately choose to use Adobe Firefly that exclusively trained on licensed materials. However, it remains difficult to ascertain whether such models have ever breached these boundaries, as it's still uncharted territory. Therefore, I will refrain from employing the AI model for creating images with artistic processing, utilizing it solely for the generation of photo-realistic visuals in this specific project where the main focus is not visual representation. I am less concerned about Co-pilot as it is still mostly a "co-programming pair" that corrects codes or provides a more efficient solution. I have to come up with the structure of the website and how each component functions myself.