Trigger words: does not need any but was trained with the word act1vate! at the start, to get as close to training data intended style as possible, include it. recommended strength: 0.9-1 or higher, lower just makes most images look like a sketch but play with settings to get a feel for it.

Additional usage tips: After further testing I find using words like "a scene from a screen play", "This photograph shows" and other indications of real-life scenes, helps preventing the images looking like sketches.

Using terminology like "background made of cardboard", "warped and deformed stage set" helps recreate the stage backdrops in the correct style.

See my example images for full prompt examples and how to control the model.

You should use act1vate! to trigger the image generation.

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.