#How to use ChatGPT/GPT4 to generate longform output

1 messages · Page 1 of 1 (latest)

coarse ocean
#

I am trying to make software that is similar to autoblogging.ai or koala.sh, that generates longform outputs from GPT4-3.5 via API, however I am struggling. Does anyone have any tips on how to go about this from a technical level?

shut onyx
#

How long output are you expecting?

#

In openai, 16~32k is max at the moment

#

If it is over than 2000 words, please try to use gpt-4-32k model

coarse ocean
#

I am expecting up to 2-3k words. I do not want to just make one API call and have it output a single long response. I do not believe this is how the major blogging tools work - they do something more clever

dry lantern
#

I assume that they simply just generate chunk after chunk

#

Generate one paragraph, integrate it into the blog and page, feed the entire blog/page back into GPT and ask it to generate another chunk/paragraph