A model creating ink an watercolor images in a western comic style. The model was trained with images of the comic series Blacksad, but it seems like Krea 2 is not all that flexible when it comes to generating anthropomorphic characters (or I did do something wrong during training), and furry characters generated with the Lora don't really look anything like in the comics. The Lora still captures the overall art style though.The model was trained without any trigger, but certain prompts cause it to generate photorealistic results that don't fit the style of the lora at all. Using the simple prompt like 'A drawing' seems to fix this issue. If it still looks off, you can try 'A watercolor drawing in a western comic book style.' The only time I had to use this so far was for a prompt featuring a painter painting a model in his art studio - which kept turning out more like a classical painting than the comic style the lora is supposed to produce - most likely because all the training data featuring such a scenario had been classical paintings as well.The example images where generated using <a target="_blank" rel="ugc" href="https://civarchive.com/models/958009/redcraft-or-2-or-3-int8int4fp8-scaled">Redcraft</a> but it seems to work fine with the base model as well.

Files with SHA256: 2120bdb117098a4bb4057702a4ec32694f28941c605117d6eb60ccd9d0699e77