This is a personal side project I started mainly to improve my testing methods for Lora collisions for filter out overfitting epochs more reliably.
The concept is simple:
you have the ability to make multiple layers, each one controls a Lora or an attribute that moves from start to end each frame by a value of step size.
There are two types of layers: Prompt layers (like Lora layers) where you can have multiple ones at the same time, and Attribute layers (like a clip skip) where you can have only one for each parameter type.
Tips:
Install by moving it to the /Scripts folder.
The resulting GIF is not shown after generation; instead, you can find it in Outputs/txt2img-images/txt2gif.
When adding a Lora layer, just input its name (e.g., "lora:Steampunk:1"), which becomes "Steampunk."
This script is still in the making, so stuff will most definitely break or have bugs.
Description
FAQ
Comments (12)
Thanks for the script, nice one!
Would be good also to have gif quality slider or smth, it made it too compressed for me so i had to animate sequence myself with another software.
I'll try to address that in the next iteration.
difficult to maneuver, how'd you suggest to prompt some coherence?
i got "ValueError: max() arg is an empty sequence", what to do?
I need bit more context, dont you know how to reproduce this bug ?
when running this with multiple batches or rows the output appears to be getting saved to a grid of a single png rather than separately. Do you have any advice/does this ring a bell?
I've experienced this with other gif making scripts which makes me wonder if maybe it's something with diffusion itself and not the scripts leveraging it.
well the script work with the 1st resulted image of each processed quarry, and when running batches that 1st result is a grid of batches results.
@navimixu yeah, i'll take a look at it when i have the spare time and see if there isn't an easy way to tweak it to do so. Was just wondering if there was a setting or something I was missing initially.
Can you add denoising parameter to this for img2img?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.

