<h1 id="dune-part:-two-cinematic-style-jfmq4qd4d">Dune Part: Two Cinematic Style</h1>Pony style LoRA that improve colors themes and background styles<h3 id="features:-x50ad5u2r">Features:</h3><ul><li>Improved Shadows</li><li>Desert, Sand and Building Themes</li><li>Muted color (low saturation)</li><li>Science fiction style</li><li>No LoRA activation trigger word BUT, I've included the tagger wordlist (Training Data file) for you to trigger specific elements from the LoRA! I absolutely love using the tag words - I hope you do, too.</li></ul><h3 id="tips:-6k0updi6y">Tips:</h3>"sand", "inside", "outside", "science fiction", and "desert" do well to generate scenes"blue eyes" or "blue" may be needed for negative or positive prompts depending on what you are doing. Training picked up blue eyes so sometimes generation will make a random item blue"scar", "helmet", "long sleeve" to prompt outfits"looking at viewer" to help look at viewer.<h3 id="training:-op5pqr36y">Training:</h3>This LoRA is trained on Dune Part: Two scene and background images with a focus on the desert scenes. The colors are muted and washed out. You may need to use "blurry", "burry background", "blurry foreground" for negative prompts because the movie has a lot of these elements naturally in all the scenes. Use negative prompt to remove if needed.Image dataset does not contain recognizable real people because of Civitai's limitation for using LoRAs to generate on-site images. This means that you can generate any character that you want. The LoRA is not biased towards people from the movie.All images were tagged using WD14. Removed and corrected minimal tags.

Dune: Part Two Cinematic Style

Trained on Cyberrealistic Pony v8. Should work with all other Pony Checkpoints.LoRA Strength: All ranges work appears to be good between 0.7 to 1. Just depends on how much influence you want from the LoRADataset Images: 147Steps: 14,600No required trigger word but use the tagger list to get specific outcomes.Filename: dunepartII_cyberrealisticv8_v1.safetensors

Files with SHA256: 21d4c9cc63bc76ebb85ca8d44235b46937b55679198253493b6c3ca43ad6b20f