"Model tilt" is a process of adding noise to a subset of the model weights to produce potentially interesting results. I see this as a parallel to cartrigde tilting, hence the name.
- Copy
model_tilt.pyto custom nodes directory in Comfy UI - Experiment away!
If done correctly, you will now see a "Model Tilt" node available under "model_tilt". This node takes a model as an input and outputs a model with applied noise.
You can tune the following parameters:
key_re: a regex for matching model keys. I personally like to tilt1.normor1.attn.strength: amount of noise to add; the noise will be generated in the[-tilt_strength .. tilt_strength]range.noise_seed: seed to use for tilt noise generation; each subsequent weight tensor will have this seed incremented internally.dropout: threshold for the mask; 1 means no effect and 0 means full effect.dropout_seed: seed to use for mask noise generation.
