This LoRA was trained on 112 images and is meant to emulate American-style cartoon images, similar to the kinds of cartoons you may have seen on TV every Saturday morning back in the day.It works really well with this checkpoint: <a rel="ugc" href="https://civarchive.com/models/7371/rev-animated">ReV Animated</a>I haven't tried it with a lot of other checkpoints, but I think it's best to use it with models or checkpoints that are already meant to produce non-photographic images.<h3>Prompting:</h3>If using <a rel="ugc" href="https://civarchive.com/models/7371/rev-animated">ReV Animated</a>, placing ((masterpiece)), ((highest quality)), and ((detailed)) in front of the prompt is useful, however you can remove ((detailed)) or lower the weight and it seems to produce flatter/simpler images.The appropriate LoRA weight seems to be between 0.4 and 0.7. At around 0.8 I notice things get a little weird. But please experiment with this. Lower weights will be subtle but still affect the output. Triggers for the LoRA are 'saturdaymorning' and 'cartoon'. You can use one of them or both of them, or neither. If you use both words, the LoRA will be stronger at lower weights. With only one of the trigger words or neither of them, you can push the weight higher.See sample images for prompting examples.And I have been using these negative embeddings:<a target="_blank" rel="ugc" href="https://huggingface.co/embed/EasyNegative/tree/main">EasyNegative</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/4629/deep-negative-v1x">Deep Negative</a><a target="_blank" rel="ugc" href="https://huggingface.co/embed/bad_prompt/blob/main/bad_prompt_version2.pt">bad_prompt_version2</a><a target="_blank" rel="ugc" href="https://huggingface.co/nick-x-hacker/bad-artist/blob/main/bad-artist.pt">bad-artist</a><a target="_blank" rel="ugc" href="https://huggingface.co/nick-x-hacker/bad-artist/blob/main/bad-artist-anime.pt">bad-artist-anime</a>

Files with SHA256: 01926b760572a7ca721857022e2d238d79f9103ce870988d494a66f09f14ee96