Users need generative AI image creation tools to accurately spell and punctuate text within images. Despite explicit instructions, current tools often ignore provided text, leading to errors. The request is for the AI to adhere to text instructions and improve overall text rendering quality.
Generative AI for image creation has shown promise, but my recent experience highlighted some significant challenges. Over the past few months, I have generated usable images using various platforms, including Poe, ChatGPT, Microsoft Co-Pilot, and Adobe Creative Suite. My most recent goal was to create an image for a LinkedIn post featuring my suggested reading list, emphasizing author names and book titles. I tried all of the above platforms. Despite writing specific prompts and revising them across several iterations, I encountered numerous issues with text in the images. While the design quality was generally acceptable, the AI struggled with accurately spelling and punctuating the text. Explicit instructions to adhere to the provided text were often ignored, leading to amusing yet frustrating results. In some cases, the platforms altered my text, adding titles and authors not included in my list while omitting others. This attempt to "improve" my list was perplexing, especially since my request was solely for an image. I encountered misspellings, misplaced quotation marks, repeated text, and missing information across all platforms. This experience reaffirmed that graphic designers are not likely to be replaced by AI anytime soon. Ultimately, I had to revert to traditional methods and design my images in Canva. What began as an effort to save time turned into a considerable investment of time and energy. To illustrate the challenges I recount above, I have included the first image that was generated by ChatGPT. The list I provided included 36 book titles and 7 podcasts. The image created to illustrate that list included 3 of the 36 book titles and none of my suggested podcasts. Instead it included two books and two podcasts I've never heard of and the icons for them are off. The good news is, I have now learned what Stoicism is and I plan to check out The Knowledge Project Podcast by Shane Parrish. It seems he has interviewed many of the people who inspire me. Share your experiences with generative AI for image creation below! #AI #generativeAI #graphicdesign #books #podcasts #chatGPT #copilot #microsoft #poe #adobe