Question about reproducing training images using respective training texts

[AhmedGamal411]

I was wondering if I had a model trained with a certain training data (text-image pairs). Would I be able to reproduce the same exact image using the the same exact text ? or will the model give another relevant image ?

For instance if I have this training pair:

source: https://imagen.research.google/

Will this input text: "A small cactus wearing a straw hat and neon sunglasses in the Sahara desert." give me the same image shown above.

If the model would give another relevant image, how can I force the model to give me an image that is as close as possible to the training image ?