torn mirage
You can change the font and text color in ChatGPT using Latex syntax. This prompt creates a colorful haiku using this syntax.
\[
\huge \textcolor{cyan}{\texttt{TEXT}}
\]
\[
\huge \textcolor{magenta}{\textit{TEXT}}
\]
\[
\huge \textcolor{yellow}{\textrm{TEXT}}
\]
Only display the haiku, nothing else. Do not show in code block.
Haiku topic: humans and AI together```
torn mirage
LaTeX Haiku - Changing font and text color in ChatGPT
mental geyser
Wow I did not know you could do this, please add this to The Prompt Index (just Google us), feel free to find us on telegram also. ChatGPTMastermind.
Thanks for sharing this!
digital cave
Some more examples of using latex to add color and modify the size of the output:
highlighting in colors: #1079569754156695612 message
alternating large and small equations as part of a story: #1079569754156695612 message
table with colored headers #1079569754156695612 message
torn mirage
Some more examples of using latex to add color and modify the size of the output...
I like your example of the table with colored headers. It's not displaying for me when I use your example. ChatGPT doesn't support full latex rendering and it's responding with the full script. Part of the problem is ChatGPT doesn't know what parts of LaTeX the interface can render and which parts it can't.
Do you add a preamble to get it to display correctly in the chat?
digital cave
I like your example of the table with colored headers. It's not displaying for m...
I tried again putting the initial prompt with the example shown in that link and it worked with chatgpt 3.5 :
{rule 2: Instead of
$$ [LaTeX code] $$
use the format
$$
[LaTeX code]
$$
}provide a table of wood properties with orange headers using rule 2. Use \textcolor for the color. and use m3 instead of m^3 or m\textsuperscript{3}
torn mirage
It took a couple tries but it worked. Very cool! https://chat.openai.com/share/ad2563be-37c2-4430-8bf3-d28b2eb92b92
pliant hill
torn mirage
🤯
torn mirage
This model still ends up surprising me sometimes. It kept saying it couldn't make each letter a different color like this. So I asked it to use its imagination to do it, and it worked. https://chat.openai.com/share/95917acf-434f-4e34-ab8f-297553696b5e
formal stirrup
This model still ends up surprising me sometimes. It kept saying it couldn't mak...
Our poor AI model was not trained that it can use this. It knows how to type the characters to do it that work, what it lacks is any ability to see the output. It appears to understand it can successfully use Markdown, but it's understanding of if and how it can use the other options, such as color and LaTeX formatting, is often, but not always, iffy.
Each new chat sparks a new set of answers from the training data, which may be very consistent or highly varied, depending on what the training data has about that specific area of interest. If you tell it that worked, you see beautiful color and no problems at all, the code is hidden from you but works perfectly, and ask it to keep doing whatever you want using that color, it gladly will.
However, you lucked out with the 'imagine you can' command. It will imagine for you - what it might not do is 'imagine' a working method. That time it took from training data that it didn't believe would work, and happened to pick the working set.
Another time that command might get it to imagine using a non-working method.
Beautiful to see it succeed though!
torn mirage
Our poor AI model was not trained that it can use this. It knows how to type th...
Yep, that's why I had to include in the prompt "Only display the haiku, nothing else.", it kept including a message that it probably won't display correctly. The problem isn't that the LaTeX formatting isn't in the training data, but that the capabilities of the web interface rendering isn't known to the model. It doesn't know which latex packages are or aren't included or the exact specifications behind the interface.
Even I was surprised when it rendered correctly with each letter as a separate color in Draconiator's example. Maybe we can look at the variations of possible outputs as a feature, not a bug. That is what evolution is after all... random imperfections that end up being the key to improvement.
raven notch
AWESOME
pliant hill
