Prompt that Imitating Someone's Writing Style adds unwanted text!

I am writing a script that gets old chats, extracts an individual’s writing style, and applies it to every new message.

Rewrite this {message} in the writing style of this {text snippets}

Doesn’t seem to work as well with short messages. It alters relevant keywords and distorts the intended meaning.

I’ve attempted to only modify specific cosmetic aspects that wouldn’t significantly affect the content like emojis and slang, yet it sometimes still adds unnecessary text. Ideas on how to improve this issue?

Change the following text according to these steps:

  • In the following text look for words or phrases that make this text sound high-level language and replace it with {medium-level language}

  • In the following text replace dots as sentence separators with new lines

  • In the following text use one of these emojis {:man_surfing: :ocean: :blush:} {one} times where it is most fits to convey emotion

  • In the following text look for verbs where it gives a command or tells the other person what to do, and replace those verbs with nicer verbs that suggest them to do it

Don’t add extra text beside the original with the changes.

Text:

Yes, the boys’ classes are conducted with a male guide, and the girls’ classes are conducted with a female guide.

Response:

1 Like

Providing examples of what kind of output you want is very beneficial and in this case, it seems turning the temperature down helps as well. I’ve also made some other changes that seem to improve performance, here’s what I ended up with after tweaking it for a while:

Change the following text according to these rules:

  1. In the following text look for words or phrases that make this text sound high-level language and replace it with words or phrases that make it sound like medium-level language.

  2. In the following text use one of these emojis {:man_surfing: :ocean: :blush:} just one time where it is most fits to convey emotion.

  3. In the following text look for verbs where it gives a command or tells the other person what to do, and replace those verbs with nicer verbs that suggest them to do it.

  4. In your response include just the new converted text without any additional information or responses. There is no need to provide any explanation or description of changes.

Here is an example of what I’m expecting:
“”"
Example Text: The proliferation of advanced machine learning models has engendered unprecedented breakthroughs in the realm of artificial intelligence.

Example Output: The spread of complex computer learning systems has led to new and exciting developments in artificial intelligence. :computer:
“”"

Now proceed with the following text.
Text:
“”"
Yes, the boys’ classes are conducted with a male guide, and the girls’ classes are conducted with a female guide.
“”"
Output:

3 Likes

This is the way to do it! Don’t tell, show.
Makes life much easier for the LLM.

1 Like

The 4th rule worked out great!
When giving it a one-shot example which starts with “Yes!” it starts to add it to many message without context… Any idea how to try to remove it?

Thanks Jack :blush:

I think I understand what you mean but I’m not positive, could you provide some examples of the text and outputs? It could just be a quick fix of modifying the example text to start with Yes! and then have the example output be formatted correctly so I’d try that too.