Generated reference · ComfyUI workflow · Flagship (frontier)

Wan 2.2 Video

Flagship 2026-generation open video on Wan 2.2 TI2V-5B: text-to-video and image-to-video in one graph (start image feeds the hybrid latent), 24fps up to 121 frames at 1280x704.

Flagship (frontier)13nodesGPU L40Sout video~240sv1.0.0
workflow.json
On this page
13Graph nodes
12Node classes
11Inputs
4Models

Tags: flagshipwan2.2videoti2v2026-sota

Inputs (11)#

The typed parameter surface callers bind when they request this workflow. Enum options and numeric bounds are the values the workflow document declares.

modeenumdefault text_to_video
Pure text-to-video, or animate a start image
options: text_to_videoimage_to_video
promptstringdefault slow aerial drift over a misty mountain temple at dawn
Motion-and-scene description in natural prose
start_imageimage
Start frame (image_to_video mode only)
resolutionenumdefault 720p_landscape
TI2V-5B native resolution (multiples of 32)
options: 720p_landscape720p_portrait
framesintegerdefault 121min 25max 121
Video length in frames (24fps; must be 4n+1 - the latent quantizes in steps of 4)
fpsintegerdefault 24min 24max 24
TI2V-5B is trained at 24fps
stepsintegerdefault 20min 10max 30
Sampling steps (official reference: 20)
guidancefloatdefault 5.0min 1.0max 10.0
CFG (official reference: 5.0)
shiftfloatdefault 8.0min 1.0max 12.0
ModelSamplingSD3 shift (official reference: 8.0)
seedintegerdefault -1
Random seed (-1 for random)
fast_modebooleandefault false
Turbo distill (Self-Forcing rank-64 LoRA): 4 steps without CFG - ~5x faster segments at slightly lower ceiling quality

ComfyUI node graph (13)#

The executable ComfyUI prompt graph: 13 nodes across 12 distinct node classes, wired by 14 data dependencies. Nodes tinted green come from a custom node pack this workflow declares; the rest are ComfyUI core / baked-community classes.

flowchart TD n0["1: UNETLoader"] n1["2: CLIPLoader"] n2["3: VAELoader"] n3["4: ModelSamplingSD3"] n4["5: CLIPTextEncode"] n5["6: CLIPTextEncode"] n6["7: LoadImage"] n7["8: Wan22ImageToVideoLatent"] n8["9: KSampler"] n9["10: VAEDecode"] n10["11: CreateVideo"] n11["12: SaveVideo"] n12["turbo_lora: LoraLoaderModelOnly"] n12 -->|model| n3 n1 -->|clip| n4 n1 -->|clip| n5 n2 -->|vae| n7 n6 -->|start_image| n7 n3 -->|model| n8 n4 -->|positive| n8 n5 -->|negative| n8 n7 -->|latent_image| n8 n8 -->|samples| n9 n2 -->|vae| n9 n9 -->|images| n10 n10 -->|video| n11 n0 -->|model| n12
Data-flow DAG — scroll to zoom, drag to pan.

Nodes (13)#

1UNETLoadercore
unet_name = wan2.2_ti2v_5B_fp16.safetensorsweight_dtype = default
outputs: MODEL
2CLIPLoadercore
clip_name = umt5_xxl_fp8_e4m3fn_scaled.safetensorstype = wan
outputs: CLIP
3VAELoadercore
vae_name = wan2.2_vae.safetensors
outputs: VAE
4ModelSamplingSD3core
model = ◂ node turbo_lora · out[0]shift = {{shift}} tmpl
outputs: MODEL
5CLIPTextEncodecore
text = {{constructed_prompt}} tmplclip = ◂ node 2 · out[0]
outputs: CONDITIONING
6CLIPTextEncodecore
text = overexposed, static motion, blurry details, subtitles, watermark, jpeg artifacts, deformed limbs, fused fingers, cluttered background, walking backwardsclip = ◂ node 2 · out[0]
outputs: CONDITIONING
7LoadImagecore
image = {{start_image}} tmpl
outputs: IMAGEMASK
8Wan22ImageToVideoLatentcore
vae = ◂ node 3 · out[0]width = {{resolution_map[resolution].width}} tmplheight = {{resolution_map[resolution].height}} tmpllength = {{frames}} tmplbatch_size = 1start_image = ◂ node 7 · out[0]
outputs: LATENT
9KSamplercore
model = ◂ node 4 · out[0]positive = ◂ node 5 · out[0]negative = ◂ node 6 · out[0]latent_image = ◂ node 8 · out[0]seed = {{seed}} tmplsteps = {{4 if fast_mode else steps}} tmplcfg = {{1.0 if fast_mode else guidance}} tmplsampler_name = uni_pcscheduler = simpledenoise = 1.0
outputs: LATENT
10VAEDecodecore
samples = ◂ node 9 · out[0]vae = ◂ node 3 · out[0]
outputs: IMAGE
11CreateVideocore
images = ◂ node 10 · out[0]fps = {{fps}} tmpl
outputs: VIDEO
12SaveVideocore
video = ◂ node 11 · out[0]filename_prefix = wan_videoformat = autocodec = auto
turbo_loraLoraLoaderModelOnlycore
model = ◂ node 1 · out[0]lora_name = Wan22_TI2V_5B_Turbo_lora_rank_64_fp16.safetensorsstrength_model = 1.0
outputs: MODEL

Prompt construction#

template
{base_prompt}. Smooth cinematic camera movement, coherent motion, rich natural lighting, filmic color.

Variables (1)#

base_prompt
{{prompt}} tmpl

Parameter banks (1)#

The prompt / configuration lookup tables this workflow keys into from its inputs — the vocabulary that turns a style / palette / preset selection into graph parameters.

resolution_map (2)#

720p_landscape
{"width": 1280, "height": 704}
720p_portrait
{"width": 704, "height": 1280}

Models & dependencies#

Models required (4)#

Wan22_TI2V_5B_Turbo_lora_rank_64_fp16.safetensors
umt5_xxl_fp8_e4m3fn_scaled.safetensors
wan2.2_ti2v_5B_fp16.safetensors
wan2.2_vae.safetensors

Output contract#

What a successful run of this workflow returns.

type
video
format
mp4

Taxonomy & routing#

How the control plane classifies this workflow — from the committed workflow-taxonomy-registry.json. It drives the consistency / control surface the agentic director can exercise over the workflow.

assetFamily
contemplative-video-bundle
outputPackageProfile
video-master-profile
controlModalities
model-locksampler-scheduler-lockseed-locktemporal-lockreference-ensemble
consistencyDimensions
environmentlightingmotion
notes
Flagship 2026 open video on Wan 2.2 TI2V-5B: t2v and start-image i2v in one graph.