Using THE IPTC ASSISTANT is intuitive and much less complex than manually editing metadata. Simply import one or more images. ChatGPT limits each request to 10 files. For optimal use, I recommend starting with requests with fewer images, associating a specific context with each image or small group (2, 3, 4 or 5 images maximum). This avoids confusion for both the assistant and you when receiving the results.
Sending too many images at once can complicate the process. For greater efficiency, focus on more targeted requests.
Also add basic information, at least the date and location. For purely illustrative images, these two elements may suffice: the assistant will analyse the visual content themselves to complete the fields.
However, they are not infallible. They cannot guess what is not visible or obvious in the image. That’s why it’s useful to enrich the context with additional details: the issue addressed, the subject of the report, or even the species of an animal or plant. Another effective method is to add a link to a Wikipedia page, the event website, or simply copy and paste the associated press release.
To save even more time, you can use the voice dictation feature (microphone icon at the bottom right) to convey your instructions and the associated context more quickly.
THE IPTC ASSISTANT can also generate IPTC fields even if you do not upload your images to the chat. To do this, it will need more detailed context and description. Based on this information, it will reformulate the data into IPTC field format, according to the instructions provided. However, it is essential to explicitly state that you will not be sending any images and that you want it to provide the description without visual support.
Limitation and solution for identifying people in images
Security rules prohibit ChatGPT from identifying or naming people in images, in particular to avoid any form of facial recognition. Sometimes, this issue applies in particular to photographs featuring political or media personalities.
To get around this limitation, the solution is to provide a broader context in THE IPTC ASSISTANT chat without sending the image. Simply describe who is in the image and specify the key elements. This will allow the tool to write the necessary metadata. In this specific case, generating IPTC fields without importing images into the chat has a practical use.
Verification and corrections
Generating IPTC fields can take a few seconds. I usually use this time to prepare my next request. Once the metadata has been generated, it is ready to be copied and pasted into Lightroom or any other image management software. In 80% of cases, the generated fields are directly usable. However, it is recommended that you check them systematically to correct any errors or add any information that the assistant was unable to deduce.
Remember, THE IPTC ASSISTANT is like an intern. It can save you a lot of time, but trust does not exclude verification.
It is also recommended that you use a different chat for each report subject.