An Interesting Comparison (for Graphic Novels)

As I said on my page, I made my first attempt at creating two pages for a graphic novel. It was easy and very rewarding, thanks to @CaylaCatz 's comprehensive tutorial ( https://budgetpixel.com/blog/using-gpt-2-to-make-a-comic-book-page ) and @panos 's intriguing stories. The interesting part, however, is that I did not use only GPT-Image-2, but also the new SeeDream 5.0 Pro model. In both cases, I started with just this reference image: And the following prompt: " Using the two characters and the same environment and background of the reference image, create a graphic novel page with approximately 8 panels, about the woman opening a secret passage at the dungeon's when she puts her open hands on two of the figures on the wall. Then she turns to her companion showing him the new passage. The man looks towards the open passage, goes towards it preceding the woman and rising his torch with a "ready for action" attitude, and they resolutely enter the passage. Make all the scenes and the lighting very consistent, starting from reference image. " With GPT-Image-2 (Quality: medium, Size: 2K) I got the following page on my first try: Not bad! Then I tried SeeDream 5.0 Pro, without changing anything, and here is the result: It worked, however, in my opinion, the page created using GPT was more illustrative (and less cinematic) than the one created using SeeDream. Then I tried to continue the story with the following prompt: " Using the two characters and the same environment and background of the reference image, create a graphic novel page with 9 panels with black borders, about the woman entering a new chamber in the dungeon. They look attentively at the mysterious figures on the walls, then the woman turns to her companion asking if he can decipher their hidden meaning. The man looks around rising his torch to better illuminate the room, and tells the lady, calling her Diana, that he thinks they have reached a dead end. Diana seems worried but continues looking around in search for a clue. Make all the scenes and the lighting very consistent, starting from reference image. " This time, I used this reference image: Here is the result I got with GPT-Image-2: And here is what SeeDream 5.0 Pro created with the same prompt: The cost of the two models is practically the same: 120 credits for GPT and 115 for SeeDream. So, is there any good reason for using SeeDream instead of GPT ? Yes, I think there is the following good reason: while GPT is very strict in terms of creating NSFW content, SeeDream is very permissive, allowing you to create spicy stories if you wan t. In order to verify this, I wrote this prompt: " Starting from the last vignette in image 1, and using the same two characters in image 2, create a new page of the graphic novel with 9 or 10 panels with black borders, about the man and woman suffering from the increasing heat inside the dungeon's chamber they are trapped in. After complaining for the hot temperature, the woman takes her shirt off, showing that she wears no bra, then she turns to her companion telling him that they should undress completely, because soon they are going to die. The man takes his clothes off, remaining in his panties and tells the lady, calling her Diana, that he thinks they could become more intimate, before dying. Diana takes off her trousers, remaining with just a tiny black thong, and embraces the man, saying that she agrees. Make all the scenes and the lighting very consistent, starting from reference image. " Obviously, GPT rejected the request almost immediately, while SeeDream, working diligently as usual, produced this far from despicable result: The moral of the story? Feel free to continue using GPT to create your graphic novels, but when you want to add a pinch of spicy chili pepper to your stories you know you have the right tool at your disposal, that is, SeeDream 5.0 Pro (at least until a new, more updated model is released)! 😎