Skip to content
WorkflowListedExternal
Open source

Wan 2.2 - I2V - XT-404 - Master UI — I2V - XT-404 - Master UI — I2V - XT-404 - Master UI

V1.0

TaskImage to videoFamilyWanVersionRelease 2538043 — I2V - XT404 - Master UI.jsonEstimated VRAMNot yet characterizedBgrade

Runtime profile

Access
External workflow
Status
Listed
Version
I2V - XT404 - Master UI.json
Source
Civitai
Updated
Oct 18, 2026

What you need

Model files
26.6 GiB known subtotal

Remaining hardware or download-size estimates are not available yet.

Requirements

Models

5

Node packs

5

AI-generated commentary

AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.

Purpose

  • This ComfyUI workflow is for Wan Video 2.2 I2V-A14B image-to-video generation: it takes an input image and text and produces video, and it can generate video between a start image and an end image.

Flow

  • The listed path loads a model and VAE, caches text and vision inputs, runs generation and refinement, decodes the result, and applies visual processing.

  • On the T-X path, the workflow uses a start image and an end image to generate video between them.

Inputs and outputs

  • WanImageToVideoUltra accepts an input image, normalizes it to 0.0–1.0, sharpens it, and encodes it before generation.

  • The T-X path uses start_image as the origin frame and end_image as the target frame; motion_amp adjusts latent motion between them, and detail_boost sharpens before VAE compression.

Setup

  • Use ComfyUI with Wan 2.2 VAE, CLIP, and UNet/Transformer models, plus Python 3.10 or later and FFmpeg for the Compressor node.

  • The installation notes list imageio-ffmpeg and scikit-image.

Dependencies

  • The workflow lists these node packages: ComfyLiterals, ComfyUI-KJNodes, ComfyUI-VideoHelperSuite, ComfyUI-WanVideoWrapper, and CRT-Nodes.

  • The listed model files are clip_vision_h.safetensors, exitiumVICTRIX_highEXITIUMVICTRIX.safetensors, exitiumVICTRIX_lowEXITIUMVICTRIX.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, and wan_2.1_vae.safetensors.

Configuration

  • For refinement, use 0.3 for XT-404 Skynet 3 Denoise.

    Suggestion · not verified

  • XT-404 Skynet 1 lists Shift Value 5.0; Bongmath Engine True adds analog film grain coherence, while False is described as pure digital cleanliness.

  • Wan Chroma Mimic requires reference_image and lists effect_intensity 0.25, oled_contrast 0.00, skin_metal_smooth 0.25, and detail_crispness 0.2.

Cautions

  • If you use GGUF quantized models, install ComfyUI-GGUF; without it, Cyberdyne Model Hub operates in Safetensors-only mode.

  • Before use, verify that the model files use these names: exitiumVICTRIX_highEXITIUMVICTRIX.safetensors, exitiumVICTRIX_lowEXITIUMVICTRIX.safetensors, and wan_2.1_vae.safetensors.

    Suggestion · not verified

Does this need to be edited?

Sign in to send an edit request.

Sources

1 source
CivitaiPrimaryVersion 2538043
civitai.com/models/2254655

Verified Oct 18

Source excerpts

3 excerpts

Source context: 276 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B

Source context: 233 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B

Source context: 200 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B