How to rewrite long articles? 2500+ words

How do you rewrite long articles? Right now i put in a prompt like 'Rewrite this in your own words make sure to use at least the same word count or more: ’

But it will stop after 100-200 words. It also will copy the article sometimes and not re-write it.


DAVINCI instruct series tend to do that. You’ll need to use original DAVINCI to get longer content. But also you’ll need to have more adjectives. It doesn’t really understand the goal if you just say “rewrite in your own words”. What is the purpose? Make it more persuasive? More diplomatic?


nossa vi está página num podcast estou louco com tanto conteúdo novo ,nunca imaginei que poderia ter tantas informações.

1 Like

Split your document up into pieces (most likely paragraphs), prompt the instruct model to rewrite each piece in the way you want it to be rewritten and then concatenate the results back into a document. This is really simple stuff you can write a script to do this for you in any language.

1 Like

i just rewrite our article and finished all. see our article here:

What programming language did you use, and did you do it as a first run written article without it looping and doing a rewrite?

Dave, I’ve hit a wall on this subject and thought maybe you might have a suggestion.

I have a python script that takes a title that’s given as a user defined variable, then writes an outline. writes an opening paragraph for the subtopics within the outline, then writes paragraphs for each of the subtopics, and then a closing paragraph to wrap up what’s been written (which by that point is long form).

The problem I’m faced with is that I want to be able to limit the repetitive keyphrases within the content by replacing them with semantic alternatives, plus I’ve noticed that every first sentence for each of the subtopic paragraphs all seem to regurgitate the subtopic title which just comes across as redundant. I’ve tried taking the path of having a final prompt that analyzes the article and then prompts the changes, but that’s just not working since it is faced with bypassing token limits as well as losing everything that’s already been written before it saves the final article to a text file.

Anything you could suggest would be incredibly appreciated.