# Bug sending prompt to gtp4o using the API, not in the browser

**URL:** <https://community.openai.com/t/bug-sending-prompt-to-gtp4o-using-the-api-not-in-the-browser/864397>\
**Category:** Prompting\
**Tags:** gpt-4, api\
**Created:** [July 12, 2024, 9:51pm UTC](https://community.openai.com/t/bug-sending-prompt-to-gtp4o-using-the-api-not-in-the-browser/864397 "2024-07-12T21:51:54Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![georgep](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/georgep/32/448530_2.png) [@georgep](https://community.openai.com/u/georgep)\
**Post date:** [July 12, 2024, 9:51pm UTC](https://community.openai.com/t/bug-sending-prompt-to-gtp4o-using-the-api-not-in-the-browser/864397/1 "2024-07-12T21:51:54Z")

</div>

I have a process where I use gpt4o via API to format and correct mistakes from GoogleAI OCR output by:

1. submitting a the prompt (vai API endpoint)
2. include the OCR output text to correct
3. include the scanned image used for the OCR process

Normally this works good, but I came across a situation where no matter how I design the prompt, after gpt comes accorss this text in the image (which I will attach):  
“One and no/100 - - - - - - - - -”  
gpt4o stops processing as instructed and repeats these dashes 1000s of times to the end of the document. I will include some screenshots so you can see the problem. I hope this is a bug that can be fixed, meanwhile, is there a work-around I can use?

I have spent hours using gpt and perplexithy revising the prompt with no success.

 ![image](https://us1.discourse-cdn.com/openai1/original/4X/2/e/5/2e5ec390a1419ebb78a915a2b58db84d0e1518fe.png)  
 ![image](https://us1.discourse-cdn.com/openai1/original/4X/d/6/5/d6566d560b7be8ed5ea91cd5e430f5594cbb0619.png)