Describing an image after I generate i

Hey Makers! :blob_wave:

I created an image - great. :white_check_mark:

I grabbed it and put it into a Google folder - all good. :white_check_mark:

Now I want to access it and create descriptive text - not sure how to do it? :question:

I cannot find the correct URL to share with the integration. I can use Claude, Perplexity, or ChatGPT since I have the api for all of them.

I am also quite open to altering the flow, since I am earning and there is VAST room for improvements!

I make the image, I get the HTTP, I create the new folder, I upload the image… now I cannot get it to describe the image.

image
That is the output after I upload the file - which does work, I can see it.

Do I have to parse the data from the HTTP module or something to see the URL?

Jinkeys, Scoob - I am kinda lost. I need a code detective! => :man_detective:

Thanks for the help!

Cheers

Chris

Howdy @Chris_K;

That last Google Drive module provides a download link via Web Content Link. However, since you’re still within the timeframe for the image at OpenAI, why not reuse that URL?

Also, couldn’t you feed the image back to OpenAI for visual processing instead of Perplexity?

Good luck,

Not Velma

Zoinks… lol - thanks for playing along @ImMichaelCannon ! I really appreciate your ideas, so I hopped on that right away.

So, it is hallucinating is the issue. Scenario runs now to completion - but the text is NOT representing the image AT ALL. >.<

Yet this is the text:
Bundle 1Collection

  • Result

The image depicts a serene beach scene at sunset. The sky is filled with vibrant hues of orange, pink, and purple, casting a warm and soothing glow. The sun is partially visible near the horizon, its setting rays reflecting off the calm sea. Silhouettes of palm trees frame the edges of the image, adding a tropical feel to the photograph. Gentle waves lap at the shore, which is dotted with smooth sand and scattered seashells. The overall atmosphere is tranquil and picturesque, capturing the peaceful beauty of a coastal evening.

Yeah, not even close at all.

This is the prompt - I kept it blazingly simple:
image

There are only 2 apps in the scenario to test, so I know for absolute certainty I have the right one. IT does fire - by misfires.

Thoughts? (and thanks for any input at all, you are very appreciated, dear Make Masters).

Cheers!

As far as I am aware the openai api is not able to fetch urls. You can use the anaylze image module for this though.

Hi @Chris_K
You can you use chat gpt "Analyze images " module.

Best regards,

Msquare Automation
Gold Partner of Make
@Msquare_Automation

1 Like

:hugs:

Thanks!!

That looks great - will try that out!

Chris

1 Like