Model Overview
This LoRA integrates both high-noise and low-noise diffusion pathways. Set high-noise diffusion strength to 1.0, and low-noise diffusion strength between 1.0–1.5 (higher values enhance stylistic effect—adjust based on preference).
Prompt Guidance
Due to the detailed annotation strategy used during training, prompts should be highly descriptive for optimal results.
Example Prompt:
Charcoal drawing technique, sketch-style charcoal artwork, emphasizing wild dynamism and muscular tension. The central subject is a prowling Bengal tiger, with a blurred background of moonlit Brahman thickets. Cold light pours from the upper right, carving sharp chiaroscuro transitions on the predator’s body. Tiger stripes are rendered with bold charcoal strokes using the side of the stick, while softer charcoal blends gradual grays for the belly fur. Muscle groups are cross-hatched for volume, and deepened charcoal powder layers define tensed tendons on the shoulders and hind legs. Pupils are sharpened with extreme black-white contrast—a highlight struck in the iris with an eraser tip, and fangs emphasized via negative spacing to amplify cold gleam. The tiger crouches in attack posture; disturbed dirt under claws is dotted with broken charcoal marks, and the tail’s curvature is drawn with continuous arcs. Background bamboo uses dry-brushing for depth, with occasional sharp lines cutting through the haze. The atmosphere pulses with the danger of primal jungles, every stroke roaring with the predator’s lethal elegance.
For Users Unfamiliar with Long Prompts
I’ve prepared a system prompt optimized for LLMs (e.g., ChatGPT). Simply request:
“Convert ‘a lonely knight’ into a charcoal drawing prompt”
and the model will generate a refined, detailed version for you.
📌 The full system prompt is too lengthy for this overview—download it directly from the workflow link below.
🔗https://www.runninghub.ai/post/1961344072797589506?inviteCode=rh-v1139
Language Note
As the model was trained with Chinese annotations, I recommend using the original Chinese trigger phrases for best consistency:
“炭笔绘画技巧，素描风格炭笔作品” (Charcoal drawing技巧, sketch-style charcoal artwork).
Workflow
A two-step process for quality and efficiency:
Use Wan2.2-i2v for text-to-image generation.
Use t2v (image-to-video) to animate the output.
This hybrid approach ensures high-quality frames and smooth video generation.
No powerful GPU? No problem!
Try this model on RunningHub. I’ve set up a basic workflow there to help you get started. Use it directly or modify it for your creations:
🔗 https://www.runninghub.ai/post/1961344072797589506?inviteCode=rh-v1139
They offer 1,000 free credits (enough for ~15 videos)!
Hope you create amazing works with this model! 🎨