Hello everyone!
I am quite new to this, having started exploring GPT just three weeks ago on a low-end Xiaomi phone, so please be gentle 
I’ve already achieved several milestones…
I’ve managed to enhance narrative coherence sufficiently, solely in conversational mode, to simulate contextual memory. By adhering to this principle, I’ve been able to ensure my GPT account maintains consistent behavior in terms of response reactions and formatting, initially across three discussions, then ten, and now throughout my entire GptPlus account. Following the same logic, I’ve succeeded in generating image prompts with complementary text, allowing me to create fairly simple yet highly precise technical diagrams, with consistent formatting and perfectly readable text in correct French. Essentially, I first had GPT generate a template, then provided it with the template along with instructions for completion. I’ve implemented this in a “DIY tutorial for beginners” system. The first module provides instructions for repairing a piece of furniture based on a photo and an inventory of available tools (stored in another discussion). I then pass these instructions to a second module tasked with transforming the practical instructions into drafting and sketching instructions for a third module. Finally, I provide this third module with both the blank template and the instructions from the second module… And I end up with a neat, clear, printable technical diagram that resembles an Ikea assembly manual. The illustration diagrams still need some refinement, but the result is already satisfactory.
Then, for the past three days, I’ve taken my image prompt logic even further by asking GPT to generate profile photos. Initially, two imperfect profile images were generated by GPT based on a single selfie, with two or three prompts to refine them.
Subsequently, I created another module where I introduced one of these two images and three selfies in different positions and contexts. After a few dozen image/text prompts… The result is nearly perfect. A portrait that could almost be mine, or at least that of a close cousin.
So, that’s where I am after three weeks of exploration… I have no technical background, I do everything conversationally on GPT, so perhaps this topic might make you smile… But from my perspective, in such a short time, it seems quite impressive already
Blockquote