[Question] controle_after_generate random2fixed seed, why ?

[dnl13]

I ask myself: why is that?
When I switch from random seed to fixed seed, exactly one queue remains and ruins my last image.
I have found a lot of workarounds, for example with "global seed node" from Inspire or ksampler from efficiency-nodes or even the temp folder or history.

but now from the product design point of view: when i switch to fixed, i assume that the seed is fixed at that moment or not? similar to the behavior as it is in a1111. you set the seed from random to fixed and the just generated image is shown again.
If I for example look at a model all the time via preview node, then found a great result, then set the seed to fixed to attach a save node, it would be expected that this image is then also saved and not another is created and the seed is perhaps lost.
do you know what i mean ? is this a per design question, so is this intentional or just technically not so easy to solve, or maybe even just a bug ?
i hope you can enlighten me on this
thank you and sorry for my maybe very stupid question.

[jn-jairo]

That was discussed before on many issues/PR, let me try to explain, the idea is, if the seed is set to random, when you queue a prompt it uses the seed that is on the input for that queued prompt and generates a new one for the next prompt, so after you click to generate what you see is not what is been generated but a new one, that follows the concept that you can edit the workflow to make your next prompt while the previous one is been generated, so if you liked a previous generated prompt and you want to generate another using the same seed you need to drag that image to the canvas to load the workflow used for that image.

Other things follow that same concept like the dynamic prompts using {|} like "picture of a {dog|cat}" when you queue the prompt it will get one of dog or cat randomly and use in the image, so when you load the image to generate the same image the prompt will be the one chosen before, with just dog or cat like "picture of a dog" so you can generate the same image without random values.

So just keep in mind that what you see is not what is been generated on the queue, the current content of the workflow and what is been generated are two different things.

[dnl13]

Thank you for taking the time to explain it again. I am very grateful for that. Ok I understand. So it is a "product design decision" It seems unfortunately a bit unintuitive but well then I will have to adapt my way of working accordingly.
Thanks again and sorry I could have searched a bit more but there were so many entries for "seed" and none for "control_after_generate".
Thank you