geceff/Wan2.2-Custom-Models-GGUF

Wan2.2 Custom GGUF (Tesla T4 Optimized)

This repository provides highly optimized Wan2.2 Image-to-Video (I2V) GGUF+LIGHTNINGV2 and custom models. These variants are fine-tuned for running efficiently on memory-constrained environments, such as Google Colab equipped with an NVIDIA Tesla T4 GPU.

---

⚡ Optimal Settings for ComfyUI

To achieve perfect video motion without artifacts or image degradation (preventing fried or oversaturated visuals), we strongly recommend using the following parameters:

Parameter	Recommended Value	Note
Sampling Steps	4	When using Wan2.2 Lightning / Distilled V2
CFG Scale	1.0	Crucial for preventing burnt images
High Noise Steps	2 or 3	To lock in strong motion and structure before the Lightning layer clears noise
Low Noise Steps	3 (End Step: 4)	Fine-tuning phase
Sampler / Scheduler	euler + simple	Standard diffusion setup

👑 Note for Higher Quality (Hybrid Workflow & Hardware Restrictions):

If you want to achieve higher visual fidelity and enhance micro-details, adopting a hybrid multi-pass approach is highly recommended. This strategy significantly sharpens fine details, effectively eliminates motion blur, and prevents fried visuals.

However, due to severe hardware VRAM limitations and Web GUI overhead, you MUST strictly adhere to the following setup configurations based on your execution environment:

💻 1. Via ComfyUI GUI (Web Interface Setup)

• 🟢 NVIDIA L4 (24GB VRAM) or higher: You can comfortably run high-tier configurations via the Web GUI with these setup options:
  • Standard High-Quality Setup: Use Q8_H (High Noise) + Q8_H (Low Noise).
  • Maximum Fidelity Option: Use Q8_H (High Noise) + wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors as the final step to achieve ultimate sharpness and micro-details.
• ⚠️ NVIDIA Tesla T4 (15GB VRAM - Free Tier GUI Limits): DO NOT use the fp8_scaled model or any configurations higher than Q6_K here! Because the Web GUI consumes a significant amount of VRAM for running its interface, the available memory is extremely limited. Forcing higher settings will result in an immediate OOM (Out of Memory) crash. To remain perfectly safe, your setup is strictly limited to:
  • Baseline Balanced Setup: Use Q4_K_M (High Noise) + Q4_K_M (Low Noise).
  • Maximum GUI Ceiling: You can scale up to Q6_K (Low Noise) at most.

🚀 2. Via Backdoor (Direct Code / Colab Forms Setup)

• 🔥 NVIDIA Tesla T4 (15GB VRAM - Unlocking Full Potential): By executing via the backend script directly, you bypass the heavy Web GUI memory overhead entirely, allowing you to forcefully squeeze maximum performance out of your T4 GPU!
• The T4 Backdoor Formulas:
  • Ultimate Quality Setup: You can successfully execute the top-tier hybrid workflow: Q8_H (High Noise) + Q8_H (Low Noise).
  • Pro Option for Speed: If you want faster generation times with a minor trade-off, switch to Q6_K (High Noise) + Q6_K (Low Noise) or Q6_K (High Noise) + Q8_H (Low Noise). This delivers optimized speed while maintaining excellent visual quality compared to full high-quants.

---

💾 Available Model Variants

Choose the right variant based on your creative workflow and VRAM configuration:

🔥 High Noise Models (wan2.2_i2v_high_noise_...)

• Best for: Creative, high-motion generation, and diverse camera movements.
• Available Quantizations: Q4_K_M, Q6_K_L, Q6_K, Q8_H

❄️ Low Noise Models (wan2.2_i2v_low_noise_...)

• Best for: High fidelity, generation stability, and strictly adhering to the prompt or structural layout of your starting frame.
• Available Quantizations: Q4_K_M, Q6_K_L, Q6_K, Q8_H, and fp8_scaled

---