#How to give camera directions ??
1 messages · Page 1 of 1 (latest)
Don't use DALL-E as a camera, it's not a camera. With that said, I would recomend to use point of view insteaed
Focus on what you want to see and now what you want to see it through
Whaaat ?? nonsense ! This is exactly what I want to see and I tell it so because it's a program that takes PROMPS ! I challenge you to get anything wider than a bust shot on a wide image
You have an example prompt?, I'd gladly take it and see what could be changed
@sonic temple Are these ok?
This was my prompt: A wide image of a Brazilian woman with proper beach swimsuit, captured from the waist up enjoying the beach. The background should be blurry.
Why does it work ? I used the same wording "captured from the waist up" and it failed as the screenshots showed. I understand what you mean earlier, that I have to describe things like a belt or shoes for the model to show the parts where those belong.
You used DALL-E as camera, you want a shot. That alone adds hallucinations to the dataset
You have to tell dall-e you want to see this, from here to there, and all that
It's DALL-E by chatgpt 🤷♂️ I have personal instructions, the about me thing and how I want gpt to behave, but nothing in there about images
that affects also how DALL-E custom GPT from OAI works, try without instructions, disable them and see if that makes a difference
I deactivated the personal instructions for new chats and asked dalle what it knew about me (nothing)
So it seems to work ! thanks. Can you recommend a guide ? I've been using it since it's release but I'm wasting a lot time on this kind of issue
There's no one guide to be honest. What I would recommend is to engage in #prompt-engineering for how to do prompts better. Take part in the #daily-theme challenges. You can learn a lot there from others. And discuss problems with your prompts in #images-canvas .
Also, keep up to date with the documentation OAI provides.