My first LoRA - an Initial attempt at recreating my preferred style for vintage comic art.
Trained on nearly 400 images that were originally created in DALL-E and personally edited in Photoshop.
You may get minor signature artifacts in the corners from me forgetting to take out my artist watermark in some of the training images. I will be working to address that in v2 along with using v1 to create additional training images.
Personal Default Negative Prompt: ugly, deformed, distorted, worst quality, low quality, low resolution, sketch artifacts, bad anatomy, signature, artist name, watermark, border, panel frame, gutter, source_anime, photograph, animation, cartoon
If you like this LoRA and style of artwork, please check out my DeviantArt profile for more!
WEIGHT: 0.9 - 1.1
CHECKPOINT: I highly recommend using the Illustrious version of this LoRA for best results. My preferred default checkpoint is RouWei Epsilon.
VAE: sdxl_vae.safetensors
CLIP SKIP: I default to 2
SAMPLER: DPM++ SDE (also successfully tested w/ Euler a)
SCHEDULER: Karras
STEPS: 20-25 should be plenty
RECOMMENDED RESOLUTIONS:
SQUARE: 1024x1024
PORTRAIT: 832x1216
LANDSCAPE: 1216x832
CFG: Lower seems to be better; I default to 3.5
ADETAILER: Yes, I have been using custom ComfyUI detailing workflows when needed with no issues.
HI-RES FIX: I've been using 1x upscale with ESRGAN_4x at 0.5 denoising strength.
UPSCALE: I use 4xUltraSharp
Initial version
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights.