tyDiffusion auto render sequence of different models/settings for testing

[alexp]

in A111 there is the ability to run an X/Y chart of various parameters and generate images for each combo automatically with a press of one button.

Im not familiar with comfyui at all, but something like this would be great within tyDiffusion.

For example, running the same prompt and settings through multiple models in sequence to then review and decide which output is preferred, or testing multiple LORAs, testing different samplers, steps etc.

Basically just queuing up a sequence of generations where 1 or 2 parameters change in sequence.

[alexp]

Oops
This is the "Batch Generation" tab right !

But would be nice to be able to auto populate it

[tyFlow]

All tyDiffusion settings/functions are exposed to MAXScript - so you can easily use a script to generate all kinds of per-parameter variations/combinations. Check the docs for more info:

https://docs.tyflow.com/tyflow_AI/tyDiff...MAXScript/

[alexp]

Thanks for this pointer, I managed to get chatgpt to write what I wanted as a result.

Im sharing this below in case anyone else finds it useful. Obviously change the paths, and this is based on excluding any XL models as Im sticking to 1.5 currently for animatediff.

-- Define the folder where your models are stored
modelFolder = "E:/Models/Stable-diffusion/"

-- Get the current scene name (without path)
sceneName = getFilenameFile(maxFileName)

-- Get all .safetensors files in the folder, excluding those containing "XL"
models = getFiles (modelFolder + "*.safetensors")
filteredModels = for m in models where (findString m "XL" == undefined) collect m

-- Loop through each filtered model file
for m in filteredModels do (
-- Extract just the model filename
modelName = getFilenameFile(m) + ".safetensors"

-- Set the current model in tyDiffusion
tyDiffusion_setString "generateModel" modelName

-- Trigger the generation process (replace with actual function if needed)
tyDiffusion_generate()

-- Define the save path using the E:/GeneratedImages directory, including scene name
savePath = "E:/GeneratedImages/" + sceneName + "_" + modelName + ".png"

-- Save the output image from viewport 0
tyDiffusion_saveImage 0 savePath
)