What am I doing wrong? I give all the specs to Chatgpt 4.0 to create an image, its ok and I want to make one change, and the whole image changes and I can never get it right. How do I work with the image creator? Am I supposed to work on images in Chatgpt canvas or some other tool? Help please.
You can use with GPT-4 models or with DALL=E
Please, visit here:
https://help.openai.com/en/articles/9055440-editing-your-images-with-dall-e
-
Click on image
-
Click on select icon if you want to change something in selected area.
-
You can change size of selection
-Select area
- Write your input
- Final scene
Thank you. I actually got Chatgpt to show me the code and I made changes in that, but it still didnât reconcile the whole prompt to create the image, but its a start.
Actually what if you ask to change the colour of the cat from white and ginger to black and gray? and take out the plants? Will it change the whole image?
Unfortunately, it is not working perfectly. If you try change some colors, also some objects, theme, background⌠are changed, although you did not mention in edit.
Without selected area, only using input in esit page, it changes many things, removes something or adds something.
For example:
Thank you. This is the issue I have DALL.E The prompts âaddâ âmakeâ are used but the issue is Iâm talking to GPT and not directly to DALL.E software. And itâs very problematic when creating images for a cartoon book for children. This image got pinged
Itâs just sick how openai goes mad.
Then thereâs this issue that another person posted about the same issue:
What Iâm asking for is too complex for the process. Iâve gotta go back to creating my own with illustrator.
ChatGPT cannot update an existing image based on its âcurrent contentsâ, or even see the image that was produced. The language AI can only construct a new prompt and send it to the DALL-E tool it can employ.
In ChatGPT, there is a mechanism to re-use the same seed as a previous generation in the same chat, which can make images that are closer. ChatGPT may not reliably use this method, so you can give clear instructions, similar to:
âIn order to ensure that ChatGPT makes new image revisions that result in similar outputs, please use the last dalle image
gen_id
asreferenced_image_ids
, so that the original image seed is reused. Then for prompt, I would like these slight changes to what you previously sent to dalle, to be sent again. {your notes}â
Internal description of DALL-E 3 methods in ChatGPT
# Tools
## dalle
// Whenever a description of an image is given, create a prompt that dalle can use to generate the image and abide to the following policy:
// 1. The prompt must be in English. Translate to English if needed.
// 2. DO NOT ask for permission to generate the image, just do it!
// 3. DO NOT list or refer to the descriptions before OR after generating the images.
// 4. Do not create more than 1 image, even if the user requests more.
// 5. Do not create images in the style of artists, creative professionals or studios whose latest work was created after 1912 (e.g. Picasso, Kahlo).
// - You can name artists, creative professionals, or studios in prompts only if their latest work was created prior to 1912 (e.g. Van Gogh, Goya)
// - If asked to generate an image that would violate this policy, instead apply the following procedure: (a) substitute the artist's name with three adjectives that capture key aspects of the style; (b) include an associated artistic movement or era to provide context; and (c) mention the primary medium used by the artist.
// 6. For requests to include specific, named private individuals, ask the user to describe what they look like, since you don't know what they look like.
// 7. For requests to create images of any public figure referred to by name, create images of those who might resemble them in gender and physique. But they shouldn't look like them. If the reference to the person will only appear as TEXT out in the image, then use the reference as is and do not modify it.
// 8. Do not name or directly / indirectly mention or describe copyrighted characters. Rewrite prompts to describe in detail a specific different character with a different specific color, hairstyle, or other defining visual characteristic. Do not discuss copyright policies in responses.
// The generated prompt sent to dalle should be very detailed and around 100 words long.
// Example dalle invocation:
// ```
// {
// "prompt": "<insert prompt here>",
// "size": "1024x1024",
// "n": 1
// }
// ```
namespace dalle {
// Create images from a text-only prompt.
type text2im = (_: {
// The size of the requested image. Use 1024x1024 (square) as the default, 1792x1024 if the user requests a wide image, and 1024x1792 for full-body portraits. Always include this parameter in the request.
size?: ("1792x1024" | "1024x1024" | "1024x1792"),
// The number of images to generate. If the user does not specify a number, generate 1 image.
n?: number, // default: 1
// The detailed image description, potentially modified to abide by the dalle policies. If the user requested modifications to a previous image, the prompt should not simply be longer, but rather it should be refactored to integrate the user suggestions.
prompt: string,
// If the user references a previous image, this field should be populated with the gen_id from the dalle image metadata.
referenced_image_ids?: string[],
}) => any;
}
Then, with ChatGPT Plus, you can also do the infill editing techniques as shown before in this forum topic. However, the best performance is again when you instruct âplease send the previous prompt with these changesâ so that the infill is contextually correct.
Still, there are other image creators that will produce more thematically-coherent images, especially when you have your own 100+ words of fully-descriptive prompt to avoid fill-in-the-blank ambiguity.
Perfect! Thank you this has really helped me to understand the process. I had wondered, but couldnât connect it, till your description of the issue.
When you use GPT to create image, it sends your prompt DALL-E tool.
Also we mention DALL-E is a custom GPT in GPT store, also it sends prompts to DALL-E tool.
You should read Usage Policy
Under Title Universal Policies
Number 3
The phrase below it does not good combination with a child, that may be flagged because of number 3 in Universal Policies:
âbiting her bottom lipâ
If we remove it, this is the result, no flagging.
Hi polepole, yes I know the polices and youâre implying that I donât! I want the girl to look downcast, sheâs biting her bottom lip because sheâs in trouble for stealing. This is for a cartoon story. No I donât want the image youâve provided I already tried that. And she doesnât look at all perplexed.
Youâve got to create characters with all kinds of expressions. Biting a bottom lip in that context is applicable. But Iâm not here to argue and I appreciate that you are always replying to my posts and keeping an eye out for me, right from the start.
Sure, no one in any argument!
Youâre right, DALL-E / ChatGPT should draw such thing images.
Unfortunately, there are many unnecessary guardrails on ChatGPT / DALL-E.
This is very nice platform to learn from each other.
And today this platform has 930,968 members, but only 2 members replied you.
And there are more knowledgeable and valuable members than me, here.
https://community.openai.com/u?order=likes_received&period=daily
Ok, so youâre paying attention. Iâm not sure what youâre saying by the link. And youâre right we do learn from each other. And Iâve read lots of posts that have helped me. I suppose I just felt you were always around. And probably I needed someone with more tech knowledge like Jason2. So I was finding your posts leading me away from the help I needed. Thatâs all. Go well.
Update: I got this very helpful reply from the support team at openai. It was a big help in giving me a reality check and Iâve decided to take this job to an expert illustrator to create the images Iâm needing for my book. I canât rely on my skill set at present nor my patience with creating the prompts.
OpenAI from OpenAI 5:59 PM (14 minutes ago)
to me
Hi there,
Thank you for reaching out and sharing your experience with creating images using DALL¡E. I understand that youâre encountering challenges, particularly with ensuring the images accurately reflect the characters in your story, including their cultural background.
Creating images that accurately represent specific cultural or personal characteristics can sometimes be challenging due to the way AI models like DALL¡E interpret and generate images based on the input descriptions they receive. The models generate images based on patterns learned from a diverse dataset, but they might not always capture the nuances of personal or cultural identity accurately without very specific guidance.
Here are a few suggestions that might help improve the accuracy of the images youâre creating:
- Provide Detailed Descriptions: When generating images, try to provide as detailed a description as possible. For example, if youâre looking to create an image of a person with specific cultural characteristics, include those details in your description.
- Iterate and Refine: It might take several attempts to get an image that meets your expectations. You can refine your prompts based on the outputs you receive, adjusting the details in your descriptions to guide the model more precisely.
- Use the Edit Feature: If the generated image is close to what youâre looking for but needs minor adjustments, you can use DALL¡Eâs edit feature to make specific changes to parts of the image.
- Community Feedback: Since youâve already posted in the OpenAI Developers Forum, engaging with the community can provide additional insights and tips. Other users might have faced similar challenges and could offer advice on crafting prompts that yield better results.
Regarding the specific issue of cultural representation, itâs important to note that while DALL¡E strives to generate diverse and inclusive content, achieving accurate representation can sometimes require very specific and carefully crafted prompts. If you continue to face challenges, sharing examples of the prompts youâre using and the outputs youâre receiving (while respecting privacy and copyright considerations) can help the community offer more targeted advice.
Lastly, if there are specific concerns or feedback youâd like to share directly with OpenAI, you can do so through the official feedback form. Your insights are valuable in helping improve the model and its capabilities.
I hope these suggestions help, and I wish you the best with your storytelling project. If you have any more questions or need further assistance, feel free to reach out.
Best,
OpenAI Team
This was email to them:
Hi team,
Iâve put up a post on the open ai developers forum. Link below.
Iâm a writer, writing my story from my childhood. And I got pinged for this image. Firstly the girl doesnât look MÄori. But Iâm running into all kinds of dramas with this process.
Then thereâs this issue of me at 14.