Our organization is looking into getting ChatGPT Teams in order to create custom GPTs for internal purposes. We already know that if you upload a PDF with text and images, it only references the text and not the image.
But, if we upload an image directly, can it reference that image? Can it also return that image to the user in a response if the question applies to that image?
I did search beforehand, but any relevant posts (that I found) were over a year old and I’m sure a lot has changed in a year.
Thank you for your response and for the warm welcome
Is there any way to have it reference a link to a Google doc that has the image? I’m trying to come up with a workaround. Is it also the same for uploaded videos?
Yes you can have a text image read and have an image downloadable from knowledge like any knowledge but you cannot have it appear in a gpt session… @_j@polepole
IMO user asked if they can retrieve an image in gpt not download it to a separate system …if code analysis is on you can download any onboard knowledge and I think they need more than 20 download images max…
I mean if they wanted it downloaded they could skip it and do it like a turn the page record “ ok open image y” and you find it and click it skip the download…
I’ve met all the qualifications set out: here 120 images to a GPT. Providing any image on demand as a GPT function (and it also was instructed to use coded randomization normally). And displaying:
It is merely developing the framework where 100 images that are about your knowledge can as easily be replayed in a GPT to a user. You just need to up your GPT developer game to make it happen.
A single uploaded zip file in a GPT lets the script the AI writes unzip many files from one to the code interpreter mount point for its notebook. This can include a hierarchy, a mapping of image file name or index to information it presents or the placeholder in file search extracted text that refers to the image (because you did your own image extraction beforehand to prepare enhanced knowledge), to then be auto-displayed by Python.
The only limitation is the current GPT AI model’s ability to keep paying attention to more complex instructions.
The file contents are not “knowledge” - they are in the code interpreter execution environment. You have to make your GPT with a decent understanding of what the Python notebook can do for you and your user.
