screenshot is not entirely processed
When executing screenshot to html (which is an awesome tools everybody is looking for), from this
I got this
Any idea what s going on ?
Yeah this is a common issue. GPT Vision can sometimes be lazy. I would retry because some outputs will be better than others.
Also think when they do release a non-preview model, it'll be easier to work around these issues with prompting. Right now the model seems to change on a daily basis.
Do you have a timeline for when they do release a non-preview model ?
No OpenAI hasn't communicated that but my guess is a few weeks.
The result is a lot better with GPT4o. We continue to improve the quality as well so I'm closing this specific older issue