Runs entirely in VRAM with no memory swapping.
Wan 2.1 14B: High-Fidelity Image-to-Video with TeaCache & GGUF Quantization
Production-grade ComfyUI Image-to-Video workflow running Wan 2.1 14B on RTX 4070 (12GB VRAM). Combines Q4_K_M GGUF, UMT5-XXL FP8 text encoder, and TeaCache temporal acceleration for fluid 16fps cinematic video in under 2 minutes.
Reproducible ComfyUI workflow for Wan 2.1 14B: High-Fidelity Image-to-Video with TeaCache & GGUF Quantization using Wan 2.1 14B I2V (Flow Matching DiT) at 832×480 resolution. Requires minimum 12GB VRAM with sampler uni_pc and scheduler simple (12 steps). Includes 1-click terminal model sync and canvas JSON graph.
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/wan-2-1-i2v-cinematic.sh | bash Loading bash setup script... Loading PowerShell setup script... import modal
app = modal.App("comfyui-wan-2-1-i2v-cinematic")
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 2.1 14B I2V (Flow Matching DiT)
print("Executing Wan 2.1 14B: High-Fidelity Image-to-Video with TeaCache & GGUF Quantization on ephemeral A10G GPU...")
return {"status": "success", "slug": "wan-2-1-i2v-cinematic"}
runpodctl create pod \
--name "comfy-wan-2-1-i2v-cinematic" \
--gpu-type "NVIDIA RTX 4090" \
--image "runpod/comfyui:latest" \
--volume-in-gb 50 \
--ports "8188/http" # No external weights required Positive Prompt
Negative Prompt
LoRA Adapter Stack 1 Adapters
RTX 4070 Calibrated WeightsRequired Models 0 Models
Field Notes RTX 4070 Benchmark
Frequently Asked Questions FAQ
What GPU and VRAM are required to run Wan 2.1 14B: High-Fidelity Image-to-Video with TeaCache & GGUF Quantization?
This workflow requires a minimum of 12GB VRAM (recommended 16GB VRAM). Tested and verified on NVIDIA GeForce RTX 4070 (12GB VRAM) at 832x480 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.