Closed hobojoker closed 1 month ago
What is your wildcard prompt?
My prompt is just a wildcard file I made for the image subject "subjects" is the entire content of the impact prompt.
It's just text file with a ton of different characters (again, over 1900 lines)
On Wednesday, May 29, 2024 at 09:47:43 PM EDT, Dr.Lt.Data ***@***.***> wrote:
What is your wildcard prompt?
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>
My prompt is just a wildcard file I made for the image subject "subjects" is the entire content of the impact prompt. It's just text file with a ton of different characters (again, over 1900 lines) On Wednesday, May 29, 2024 at 09:47:43 PM EDT, Dr.Lt.Data @.> wrote: What is your wildcard prompt? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>
__subject__
wildcard file nested?The test results indicate that it's just a matter of bad luck. By dividing the seed into consecutive groups of 100 (0-100, 100-200, etc.) and testing, the collision probability ranges from 0% to 5%.
One reason is that your prompt is too simple. Even if duplication occurs with the first prompt, the probability of duplication occurring consecutively with the second wildcard in a prompt decreases exponentially.
Okay, thanks for confirming that!
It feels like there is something wrong with the wildcard seed assignment - I have a wildcard list with 1947 lines, and yet in 20 image generations with random seeds I somehow have 4 repeat images? I'm consistently getting repeats way more often with unique seed values than expected.
The 4 repeats in 20 generations is just one example, I've run hundreds of images through and I see a ton more repeats than expected, and specific ones are repeated more often than others.
Obviously there are going to be repeats, and I expect that, but it seems to repeat far too consistently...
Seed 609318467621642 produced the same wildcard output as seed 16860090012894, in 4 generations, there was a repeat!