Is it possible to have zero latent noise or choose the custom latent noise?

[Creative-comfyUI]

If I am right, if you have zero conditioning for negatif and positive prompt, the image you get is the conversion of the latent noise.
Noise generated from our prompt - latent noise -> vae = SDXL image

If we want to add our own latent noise, how can we do that? (this is not the same as image to image)
If this latent noise is not present, what will the prompt give us as an image?

[ltdrdata]

The prompt guides the direction of diffusion but does not add noise.

If you want to perform diffusion without adding initial noise, you can disable add_noise in KSamplerAdvanced.

In the case of a uniform latent like an empty latent, it is already close to a converged result without noise, making it difficult to introduce variations.

If you want to use latent as initial noise, refer to KSampler (Inspire) in the Inspire Pack. The noise_opt parameter serves that purpose.