In conjunction with What make effective start and end frames?, where I used an example of a short video generated using start and end frames plus a prompt, I will explain how to use seeds to generate effective start and end frames.
Working in Image Generation and using the model Stable Diffusion, I used the following prompt to generate the subsequent image:
"Generate a studio photo of an enthusiastic and ecstatic black man in his late twenties looking at the camera. He should be in front of a solid color vibrant yellow (#FFD143) backdrop and wearing a yellow sweatshirt in the same color. He is wearing glasses and has afro hair with a texture skin fade. Short beard. Brightly and artificially lit. Studio photography. Canon EOS, lens 50mm, aperture f/7."
I was happy with this image and wanted to use it as my end frame in a short clip in which the man turns from profile to head-on and smiles broadly, so I saved it to my Asset Library.
I took the seed from that image by hovering over it and clicking the three dots in the top right corner of the image and selecting 'View Provenance':
To generate my desired start frame I copied and pasted the seed value into the Generate Options panel 'Seed' input:
With the model noting my seed value I used the following prompt to generate the subsequent image:
"Generate a studio photo of an enthusiastic and ecstatic black man in his late twenties in profile. He should be in front of a solid color vibrant yellow (#FFD143) backdrop and wearing a yellow sweatshirt in the same color. He is wearing glasses and has afro hair with a texture skin fade. Brightly and artificially lit. Studio photography. Canon EOS, lens 50mm, aperture f/7."
I generated a few variations before selecting the one I wanted to use for my start frame, so I saved it to my Asset Library.
Note: not all models support seeds at all, or in the same way as Stable Diffusion.