Wan 2.1 14B: High-Fidelity Image-to-Video with TeaCache & GGUF Quantization
12st · uni_pc 832×480 12GB
H.264 Video
Generation Specs RTX 4070
Sampler uni_pc
Scheduler simple
Steps 12
CFG Scale 5
Seed 97097018447
Target VRAM 16GB
VRAM 12GB - 16GB
GPU Compatibility
Optimal Performance
Runs entirely in VRAM with no memory swapping.
Wan 2.2 / 2.1 14B I2V + LTX-Video 2.3 Audio Civitai #144554572 Min 12GB VRAM 480×832 RTX 4070 Verified
Wan 2.2 I2V & 8-LoRA Stack (White-Weapon Architecture)
Full-spectrum Image-to-Video synthesis pipeline reverse-engineered from Civitai #144554572. Daisy-chains all 8 multi-scale LoRA adapters sequentially with TeaCache acceleration on Wan 2.1/2.2 14B.
Blueprint Summary RTX 4070 Verified
Reproducible ComfyUI workflow for Civitai #144554572: Wan 2.2 I2V & 8-LoRA Stack (White-Weapon Architecture) using Wan 2.2 / 2.1 14B I2V + LTX-Video 2.3 Audio at 480×832 resolution. Requires minimum 12GB VRAM with sampler uni_pc and scheduler simple (12 steps). Includes 1-click terminal model sync and canvas JSON graph.
Pipeline Flow 20 Nodes
Open in Resolver 01 Load Models
UnetLoaderGGUF
11 loaders (DiT, CLIP, VAE)
02 Adapters
LoraLoader
8 adapters
03 Conditioning
CLIPTextEncode
2 prompt encodings
04 KSampler
KSampler
DiT latent denoising
05 Decode & Save
VAEDecode
Latent to pixel space
Execution DAG Topology Interactive Visualizer
Drag to pan · Scroll to zoom · Hover wiresModel & Asset Setup Setup Script
Run in your ComfyUI root:
curl -fsSL https://decomfy.com/api/scripts/civitai-wan-audio-v2v-sync.sh | bash Loading bash setup script... Loading PowerShell setup script... import modal
app = modal.App("comfyui-civitai-wan-audio-v2v-sync")
vol = modal.Volume.from_name("comfy-weights-cache", create_if_missing=True)
image = (
modal.Image.debian_slim(python_version="3.11")
.apt_install("git", "wget", "curl", "libgl1-mesa-glx", "libglib2.0-0")
.pip_install("torch", "torchvision", "--index-url", "https://download.pytorch.org/whl/cu124")
.pip_install("transformers", "accelerate", "safetensors", "aiohttp")
.run_commands(
"git clone https://github.com/comfyanonymous/ComfyUI.git /root/ComfyUI",
"cd /root/ComfyUI && pip install -r requirements.txt",
)
)
@app.function(
gpu="A10G",
image=image,
volumes={"/root/ComfyUI/models": vol},
timeout=900,
)
def generate():
# Headless serverless execution for Wan 2.2 / 2.1 14B I2V + LTX-Video 2.3 Audio
print("Executing Civitai #144554572: Wan 2.2 I2V & 8-LoRA Stack (White-Weapon Architecture) on ephemeral A10G GPU...")
return {"status": "success", "slug": "civitai-wan-audio-v2v-sync"}
runpodctl create pod \
--name "comfy-civitai-wan-audio-v2v-sync" \
--gpu-type "NVIDIA RTX 4090" \
--image "runpod/comfyui:latest" \
--volume-in-gb 50 \
--ports "8188/http" # ComfyUI Model Batch Ingestion for Civitai #144554572: Wan 2.2 I2V & 8-LoRA Stack (White-Weapon Architecture)
# Run with: aria2c -i models-civitai-wan-audio-v2v-sync.txt -j4 -x4
https://huggingface.co/city96/Wan2.1-I2V-14B-720P-gguf/resolve/main/wan2.1-i2v-14b-720p-Q4_K_M.gguf
dir=models/diffusion_models
out=wan2.1-i2v-14b-720p-Q4_K_M.gguf
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
dir=models/text_encoders
out=umt5_xxl_fp8_e4m3fn_scaled.safetensors
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors
dir=models/vae
out=wan_2.1_vae.safetensors
https://civitai.com/api/download/models/2361379
dir=models/loras
out=Wan_2_2_I2V_A14B_HIGH_lightx2v_4step_lora_v1030_rank_64_bf16.safetensors
https://civitai.com/api/download/models/2553271
dir=models/loras
out=DR34ML4Y_I2V_14B_LOW_V2.safetensors
https://civitai.com/api/download/models/2073605
dir=models/loras
out=NSFW-22-H-e8.safetensors
https://civitai.com/api/download/models/2265575
dir=models/loras
out=wan22-m4crom4sti4-i2v-20epoc-high-k3nk.safetensors
https://civitai.com/api/download/models/2298673
dir=models/loras
out=WAN-2.2-I2V-POV-Body-Cumshot-Pullout-HIGH-v1.safetensors
https://civitai.com/api/download/models/2116027
dir=models/loras
out=56Low noise-Cumshot Aesthetics.safetensors
https://civitai.com/api/download/models/2152516
dir=models/loras
out=jfj-deepthroat-W22-I2V-HN.safetensors
https://civitai.com/api/download/models/2121297
dir=models/loras
out=wan2.2-i2v-oral-insertion-v1.0.safetensors Positive Prompt
cinematic lookbook film, young woman lying gracefully on white silk bedding in luxury suite, soft morning sunlight, subtle breathing movement, turning head gently towards camera with calm confident smile, high-fidelity 4k texture, smooth cinematic motion
Negative Prompt
unclothed, disrobed, low quality, blurry, distorted, jitter, flickering, warped face, bad anatomy, deformed
Prompt Customization:
This workflow is pre-wired and calibrated. Swap this prompt with your own character, subject, or style without breaking the pipeline.
LoRA Adapter Stack 8 Adapters
RTX 4070 Calibrated Weights 01 Wan_2_2_I2V_A14B_HIGH_lightx2v_4step_lora_v1030_rank_64_bf16.safetensors +1
02 DR34ML4Y_I2V_14B_LOW_V2.safetensors +0.8
03 NSFW-22-H-e8.safetensors +0.8
04 wan22-m4crom4sti4-i2v-20epoc-high-k3nk.safetensors +0.7
05 WAN-2.2-I2V-POV-Body-Cumshot-Pullout-HIGH-v1.safetensors +0.7
06 56Low noise-Cumshot Aesthetics.safetensors +0.6
07 jfj-deepthroat-W22-I2V-HN.safetensors +0.6
08 wan2.2-i2v-oral-insertion-v1.0.safetensors +0.6
Required Models 11 Models
wan2.1-i2v-14b-720p-Q4_K_M.gguf Quantized Wan 2.1 14B Flow Matching Diffusion Transformer
umt5_xxl_fp8_e4m3fn_scaled.safetensors Official UMT5-XXL FP8 Scaled text encoder for Wan 2.1 / Wan 2.2
wan_2.1_vae.safetensors Official 16-channel video VAE (Wan 2.1)
Wan_2_2_I2V_A14B_HIGH_lightx2v_4step_lora_v1030_rank_64_bf16.safetensors LoRA 1: Lightning 4-Step Turbo Accelerator (Weight: 1.0)
DR34ML4Y_I2V_14B_LOW_V2.safetensors LoRA 2: DR34ML4Y Cinema Realism & Dynamic Motion (Weight: 0.8)
NSFW-22-H-e8.safetensors LoRA 3: Human Anatomy & Motion Dynamics (Weight: 0.8)
wan22-m4crom4sti4-i2v-20epoc-high-k3nk.safetensors LoRA 4: Volumetric Silhouette & Soft-Body Physics (Weight: 0.7)
WAN-2.2-I2V-POV-Body-Cumshot-Pullout-HIGH-v1.safetensors LoRA 5: Dynamic POV Dolly Zoom & Pullout (Weight: 0.7)
56Low noise-Cumshot Aesthetics.safetensors LoRA 6: Micro-Fluid Dynamics & Surface Lighting (Weight: 0.6)
jfj-deepthroat-W22-I2V-HN.safetensors LoRA 7: Head Tilt & Perspective Angle Tracker (Weight: 0.6)
wan2.2-i2v-oral-insertion-v1.0.safetensors LoRA 8: Facial Expression & Macro Profile Tracker (Weight: 0.6)
Field Notes RTX 4070 Benchmark
Benchmark: [object Object]
1. ARCHITECTURAL REASONING & 8-LORA DAISY-CHAINING:
Deconstructs Civitai #144554572 by creator kenpechi into a clean-room White-Weapon architecture. Instead of monolithic checkpoints, this recipe sequences 8 specialized adapters across a single tensor backbone: Lightning 4-step for rapid convergence, anatomical dynamics for physical realism, and multi-scale perspective sliders for focal depth.
2. CABLE ROUTING & SIGNAL TOPOLOGY:
Signal flows sequentially from UnetLoaderGGUF through TeaCache (rel_l1_thresh: 0.4) into ModelSamplingSD3 (shift: 8.0). The MODEL and CLIP streams cascade through all 8 LoraLoader nodes in series, terminating at a single deterministic Positive CLIPTextEncode wired directly to KSampler. Start frames are encoded via 16-channel video VAE in WanImageToVideo.
3. HARDWARE BENCHMARKS & CONSUMER FOOTPRINT:
Benchmarked on workstation wingpu (RTX 4070 12GB): 49 frames at 480x832 (16fps) execute in 263s under 12 steps of UniPC sampling. Peak VRAM stabilizes at 10.9 GB, providing safe headroom under the 12GB host threshold.
Frequently Asked Questions FAQ
What GPU and VRAM are required to run Civitai #144554572: Wan 2.2 I2V & 8-LoRA Stack (White-Weapon Architecture)?
This workflow requires a minimum of 12GB VRAM (recommended 16GB VRAM). Tested and verified on NVIDIA GeForce RTX 4070 (12GB VRAM) at 480x832 resolution.
How do I resolve missing custom nodes for this workflow?
You can drop the workflow JSON into our client-side Missing Node Auto-Resolver at https://decomfy.com/resolve/ to detect missing nodes and generate install commands, or run the 1-click terminal setup script provided below.
What hardware and precision are required for Wan 2.1 1.3B Video DiT?
Wan 2.1 1.3B Text-to-Video uses a flow-matching 3D diffusion transformer with UMT5-XXL text encoder. On an RTX 4070 (12GB VRAM), load the 1.3B DiT in BF16 alongside FP8-scaled UMT5 text encoder for fluid 5-second 720p generations.
Related ComfyUI Blueprints
View all 47 workflows
Wan 2.1 1.3B: Fast Local Cinematic Video on Consumer GPUs (8GB-12GB)
25st · euler 832×480 8GB
Qwen-Image-2.1 Turbo Uncensored: Official 8-Step DiT & Abliterated Vision-Language Architecture
8st · euler 1024×1024 10GB
