For very complex requests, such as writing applications with specific formal and content requirements that I need in german language, I write my prompt in English and then translate the response into german using Deepl.
My idea is that if you have higher requirements for the layout of a text, the output in English is generally preferable. Are there any experiences or assessments to the contrary?
1 Like
I think that is a sound logic, as the largest content the model has been trained on is English text, so it is likely to give a higher quality output when using it. You could also get the model to do the translation as a second step, it does translate quite well. Personally I do translations in 2 passes, one is the translation and the second pass is to tell the model that this text has been translated and to to correct it to sound like a native speaker wrote it.
2 Likes
Thank you for your helpful reply.
In fact, i had not thought of having the translation linguistically upgraded by the model again. I have no doubt that your approach will prove very useful to me.
1 Like
Nice this is good idea for the YouTube people