I’m hoping that this upcoming upgraded Memory feature (“Personalization” or whatever they end up calling it) will make it easier to generate consistent characters, especially across more than one chat session. We’ll see.
I am also trying to draw various scenes with the same character.
In my experience, you can get closer to the expected result by making the prompts that describe the character and style completely the same, distinguishing them from the scene prompts, creating one scene several times, and reusing the gen_id of the illustration you like. It seemed like it.
Specifically, within the same session, it seems that the same seed value will be assigned by giving an instruction to change the gen_id of the generated image.
Since gen_id cannot be used across sessions, we hope to be able to use gen_id across sessions.
To verify what I said, I attempted to create a GPT-based system that can consistently generate images with “style consistency” of the same character within the same conversation.
Draw (Nearly) Something
This is based on the principle that within the same conversation in GPT, using the same seed + the same prompt results in identical images. However, if there are too many requested changes, the system will fail. Of course, this approach will probably become obsolete once Sora is released. Haha.
hello everyone,
I want to create an app for this can someone tell me how can i use dall e API for Consistent Image generation?
Awesome. Can you please provide your prompt? I want to do it through api
Hey man, have you found any solutions since I’m facing the same question as yours? I’m trying to make my images consistent through DALLE API
Nope, I don’t think it is possible until open ai introduces seed.
yeah, so i have used midjourney which has a feature named "–sref " providing a really nice results of keeping image style consistent
An Indian man is sitting in his office, his age is 29, he is a journalist, he has a computer on his table, he has a note paper in his hand, he is reading that paper carefully.
Hi,
I think you are looking for ChatGPT, it can create images like the one you requested, ytou can find it over on chatgpt.com