Aesthetic Quality Modifiers - Masterpiece
Training data is a subset of all my manually rated datasets with the quality/aesthetic modifiers, including only the masterpiece tagged images.
ℹ️ LoRA work best when applied to the base models on which they are trained. Please read the About This Version on the appropriate base models, trigger usage, and workflow/training information.
Recommended prompt structure:
Positive prompt (quality tags at the start of prompt):
masterpiece, best quality, very aesthetic, {{tags}}, {{natural language}}Description
[WAN 14B] LoRA (experimental)
Trained with diffusion-pipe on Wan2.1-T2V-14B with the same (image-only) dataset as v2.3 [noobai v-pred]
Currently curating a video dataset
Video previews generated with ComfyUI_examples/wan/#text-to-video
Loading the LoRA with LoraLoaderModelOnly node and using the fp8 14B: wan2.1_t2v_14B_fp8_e4m3fn.safetensors
Higher quality previews use the full fp16 14b: wan2.1_t2v_14B_fp16.safetensors
Recommend following prompting guide for movement to avoid still images/jitter: https://www.comfyonline.app/blog/wan2-1-prompt-guide
Image previews generated with modified ComfyUI_examples/wan/#text-to-video
Setting the frame length to 1
Adding Upscaling
Better results with text-to-image than text-to-video for this version (due to training on images only)
FAQ
Comments (4)
is it fine to use v-pred version lora at eps checkpoint?
It's fine yes, but not sure if I would recommended it - you may notice a saturation and contrast increase similar to when using v-pred checkpoints
1.3B model for v2v por favor
Released a 1.3B version here: https://civitai.com/models/929497?modelVersionId=1697670
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.