nodes/kijai-ComfyUI-WanVideoWrapper
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Make start_image optional for Fun-InP workflows, fix control lora

Browse Source 
NOTE the ImageToVideo encode node inputs end up changing getting mixed up due to the change, old workflows need resetting those inputs.
This commit is contained in:
kijai
2025-03-28 13:10:52 +02:00
parent d9b1f4d1a5
commit 6334dfab84
4 changed files with 1283 additions and 1061 deletions
Download Patch File Download Diff File Expand all files Collapse all files
example_workflows/wanvideo_480p_I2V_endframe_example_01.json
+566 -423
View File
File diff suppressed because it is too large Load Diff
example_workflows/wanvideo_480p_I2V_example_02.json
+280 -218
View File
@@ -2,7 +2,7 @@
"id": "206247b6-9fec-4ed2-8927-e4f388c674d4",
"revision": 0,
"last_node_id": 67,
"last_link_id": 80,
"last_link_id": 83,
"nodes": [
{
"id": 46,
@@ -38,6 +38,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoTextEmbedBridge"
},
"widgets_values": []
@@ -74,6 +76,8 @@
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.26",
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
@@ -107,6 +111,8 @@
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.26",
"Node name for S&R": "CLIPLoader"
},
"widgets_values": [
@@ -147,6 +153,8 @@
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.26",
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
@@ -199,6 +207,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoVRAMManagement"
},
"widgets_values": [
@@ -325,6 +335,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "LoadWanVideoT5TextEncoder"
},
"widgets_values": [
@@ -345,7 +357,7 @@
],
"size": [
315,
154
178
],
"flags": {},
"order": 8,
@@ -361,6 +373,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoTeaCache"
},
"widgets_values": [
@@ -368,7 +382,8 @@
1,
-1,
"offload_device",
"true"
"true",
"e"
]
},
{
@@ -466,6 +481,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoBlockSwap"
},
"widgets_values": [
@@ -501,6 +518,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoTorchCompileSettings"
},
"widgets_values": [
@@ -544,6 +563,8 @@
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.26",
"Node name for S&R": "LoadImage"
},
"widgets_values": [
@@ -554,89 +575,6 @@
"color": "#2a363b",
"bgcolor": "#3f5159"
},
{
"id": 66,
"type": "ImageResizeKJ",
"pos": [
-175.95899963378906,
176.40057373046875
],
"size": [
315,
266
],
"flags": {},
"order": 21,
"mode": 0,
"inputs": [
{
"name": "image",
"type": "IMAGE",
"link": 72
},
{
"name": "get_image_size",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "width_input"
},
"link": null
},
{
"name": "height_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "height_input"
},
"link": null
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
73,
74
]
},
{
"name": "width",
"type": "INT",
"links": [
75
]
},
{
"name": "height",
"type": "INT",
"links": [
76
]
}
],
"properties": {
"Node name for S&R": "ImageResizeKJ"
},
"widgets_values": [
624,
624,
"lanczos",
false,
16,
0,
0,
"disabled"
]
},
{
"id": 63,
"type": "WanVideoImageToVideoEncode",
@@ -646,26 +584,27 @@
],
"size": [
352.79998779296875,
262
338
],
"flags": {},
"order": 25,
"mode": 0,
"inputs": [
{
"name": "start_image",
"type": "IMAGE",
"link": 74
},
{
"name": "vae",
"type": "WANVAE",
"link": 77
"link": 81
},
{
"name": "clip_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"link": 69
"link": 82
},
{
"name": "start_image",
"shape": 7,
"type": "IMAGE",
"link": 83
},
{
"name": "end_image",
@@ -688,6 +627,12 @@
"name": "height"
},
"link": 76
},
{
"name": "control_embeds",
"shape": 7,
"type": "WANVIDIMAGE_EMBEDS",
"link": null
}
],
"outputs": [
@@ -700,6 +645,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoImageToVideoEncode"
},
"widgets_values": [
@@ -709,47 +656,12 @@
0.030000000000000006,
1,
1,
true
true,
false
],
"color": "#2a363b",
"bgcolor": "#3f5159"
},
{
"id": 38,
"type": "WanVideoVAELoader",
"pos": [
169.25408935546875,
-322.9471740722656
],
"size": [
372.7727966308594,
82
],
"flags": {},
"order": 15,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "vae",
"type": "WANVAE",
"slot_index": 0,
"links": [
43,
77
]
}
],
"properties": {
"Node name for S&R": "WanVideoVAELoader"
},
"widgets_values": [
"wanvideo\\Wan2_1_VAE_bf16.safetensors",
"bf16"
],
"color": "#322",
"bgcolor": "#533"
},
{
"id": 59,
"type": "CLIPVisionLoader",
@@ -762,7 +674,7 @@
58
],
"flags": {},
"order": 16,
"order": 15,
"mode": 0,
"inputs": [],
"outputs": [
@@ -775,6 +687,8 @@
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.26",
"Node name for S&R": "CLIPVisionLoader"
},
"widgets_values": [
@@ -783,62 +697,6 @@
"color": "#233",
"bgcolor": "#355"
},
{
"id": 65,
"type": "WanVideoClipVisionEncode",
"pos": [
-176.9497833251953,
-167.4617462158203
],
"size": [
327.5999755859375,
242
],
"flags": {},
"order": 24,
"mode": 0,
"inputs": [
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 70
},
{
"name": "image_1",
"type": "IMAGE",
"link": 73
},
{
"name": "image_2",
"shape": 7,
"type": "IMAGE",
"link": null
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"links": [
69
]
}
],
"properties": {
"Node name for S&R": "WanVideoClipVisionEncode"
},
"widgets_values": [
1,
1,
"center",
"average",
true,
0,
0.20000000000000004
],
"color": "#233",
"bgcolor": "#355"
},
{
"id": 16,
"type": "WanVideoTextEncode",
@@ -877,6 +735,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoTextEncode"
},
"widgets_values": [
@@ -958,6 +818,12 @@
"shape": 7,
"type": "LOOPARGS",
"link": null
},
{
"name": "experimental_args",
"shape": 7,
"type": "EXPERIMENTALARGS",
"link": null
}
],
"outputs": [
@@ -971,6 +837,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoSampler"
},
"widgets_values": [
@@ -984,8 +852,7 @@
0,
1,
"",
"comfy",
""
"comfy"
]
},
{
@@ -1000,7 +867,7 @@
106
],
"flags": {},
"order": 17,
"order": 16,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1013,6 +880,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoEnhanceAVideo"
},
"widgets_values": [
@@ -1068,6 +937,8 @@
}
],
"properties": {
"cnr_id": "comfyui-videohelpersuite",
"ver": "0a75c7958fe320efcb052f1d9f8451fd20c730a8",
"Node name for S&R": "VHS_VideoCombine"
},
"widgets_values": {
@@ -1133,6 +1004,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoDecode"
},
"widgets_values": [
@@ -1197,6 +1070,8 @@
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoModelLoader"
},
"widgets_values": [
@@ -1208,6 +1083,193 @@
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 38,
"type": "WanVideoVAELoader",
"pos": [
169.25408935546875,
-322.9471740722656
],
"size": [
372.7727966308594,
82
],
"flags": {},
"order": 17,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "vae",
"type": "WANVAE",
"slot_index": 0,
"links": [
43,
81
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoVAELoader"
},
"widgets_values": [
"wanvideo\\Wan2_1_VAE_bf16.safetensors",
"bf16"
],
"color": "#322",
"bgcolor": "#533"
},
{
"id": 65,
"type": "WanVideoClipVisionEncode",
"pos": [
-176.9497833251953,
-167.4617462158203
],
"size": [
327.5999755859375,
262
],
"flags": {},
"order": 24,
"mode": 0,
"inputs": [
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 70
},
{
"name": "image_1",
"type": "IMAGE",
"link": 73
},
{
"name": "image_2",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "negative_image",
"shape": 7,
"type": "IMAGE",
"link": null
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"links": [
82
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d9b1f4d1a5aea91d101ae97a54714a5861af3f50",
"Node name for S&R": "WanVideoClipVisionEncode"
},
"widgets_values": [
1,
1,
"center",
"average",
true,
0,
0.20000000000000004
],
"color": "#233",
"bgcolor": "#355"
},
{
"id": 66,
"type": "ImageResizeKJ",
"pos": [
-175.95899963378906,
176.40057373046875
],
"size": [
315,
266
],
"flags": {},
"order": 21,
"mode": 0,
"inputs": [
{
"name": "image",
"type": "IMAGE",
"link": 72
},
{
"name": "get_image_size",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "width_input"
},
"link": null
},
{
"name": "height_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "height_input"
},
"link": null
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
73,
83
]
},
{
"name": "width",
"type": "INT",
"links": [
75
]
},
{
"name": "height",
"type": "INT",
"links": [
76
]
}
],
"properties": {
"cnr_id": "comfyui-kjnodes",
"ver": "52c2e31a903fec2dd654fb614ea82ba2757d5028",
"Node name for S&R": "ImageResizeKJ"
},
"widgets_values": [
624,
624,
"lanczos",
false,
16,
0,
0,
"disabled"
]
}
],
"links": [
@@ -1315,14 +1377,6 @@
2,
"WANVIDIMAGE_EMBEDS"
],
[
69,
65,
0,
63,
2,
"WANVIDIMAGE_CLIPEMBEDS"
],
[
70,
59,
@@ -1347,14 +1401,6 @@
1,
"IMAGE"
],
[
74,
66,
0,
63,
0,
"IMAGE"
],
[
75,
66,
@@ -1371,14 +1417,6 @@
5,
"INT"
],
[
77,
38,
0,
63,
1,
"WANVAE"
],
[
79,
22,
@@ -1394,6 +1432,30 @@
27,
4,
"FETAARGS"
],
[
81,
38,
0,
63,
0,
"WANVAE"
],
[
82,
65,
0,
63,
1,
"WANVIDIMAGE_CLIPEMBEDS"
],
[
83,
66,
0,
63,
2,
"IMAGE"
]
],
"groups": [
@@ -1414,10 +1476,10 @@
"config": {},
"extra": {
"ds": {
"scale": 0.6115909044841832,
"scale": 0.6727499949326018,
"offset": [
1066.099511487144,
1189.0841914350103
547.1803980844202,
840.3120224296222
]
},
"node_versions": {
example_workflows/wanvideo_Fun_control_example_01.json
+404 -400
View File
@@ -2,7 +2,7 @@
"id": "206247b6-9fec-4ed2-8927-e4f388c674d4",
"revision": 0,
"last_node_id": 95,
"last_link_id": 121,
"last_link_id": 124,
"nodes": [
{
"id": 46,
@@ -443,70 +443,6 @@
"color": "#233",
"bgcolor": "#355"
},
{
"id": 65,
"type": "WanVideoClipVisionEncode",
"pos": [
-1200.4986572265625,
-431.59814453125
],
"size": [
327.5999755859375,
262
],
"flags": {},
"order": 39,
"mode": 2,
"inputs": [
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 70
},
{
"name": "image_1",
"type": "IMAGE",
"link": 73
},
{
"name": "image_2",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "negative_image",
"shape": 7,
"type": "IMAGE",
"link": null
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"links": [
69
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d20baf00247fd06553fdc9253e18732244e54172",
"Node name for S&R": "WanVideoClipVisionEncode"
},
"widgets_values": [
1,
1,
"center",
"average",
true,
0,
0.20000000000000004
],
"color": "#233",
"bgcolor": "#355"
},
{
"id": 73,
"type": "DownloadAndLoadDepthAnythingV2Model",
@@ -540,107 +476,6 @@
"depth_anything_v2_vitl_fp16.safetensors"
]
},
{
"id": 66,
"type": "ImageResizeKJ",
"pos": [
-1190.589111328125,
-70.7883529663086
],
"size": [
315,
266
],
"flags": {},
"order": 36,
"mode": 2,
"inputs": [
{
"name": "image",
"type": "IMAGE",
"link": 72
},
{
"name": "get_image_size",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "width_input"
},
"link": null
},
{
"name": "height_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "height_input"
},
"link": null
},
{
"name": "width",
"type": "INT",
"widget": {
"name": "width"
},
"link": 97
},
{
"name": "height",
"type": "INT",
"widget": {
"name": "height"
},
"link": 98
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
73,
74
]
},
{
"name": "width",
"type": "INT",
"links": [
75
]
},
{
"name": "height",
"type": "INT",
"links": [
76
]
}
],
"properties": {
"cnr_id": "comfyui-kjnodes",
"ver": "52c2e31a903fec2dd654fb614ea82ba2757d5028",
"Node name for S&R": "ImageResizeKJ"
},
"widgets_values": [
624,
624,
"lanczos",
false,
16,
0,
0,
"center"
]
},
{
"id": 69,
"type": "WanVideoEmptyEmbeds",
@@ -707,29 +542,6 @@
81
]
},
{
"id": 81,
"type": "Note",
"pos": [
387.06109619140625,
44.78668975830078
],
"size": [
255.5326690673828,
88
],
"flags": {},
"order": 12,
"mode": 0,
"inputs": [],
"outputs": [],
"properties": {},
"widgets_values": [
"Use empty embeds when doing just text to video with control, otherwise connect the image embeds from the image encoder"
],
"color": "#432",
"bgcolor": "#653"
},
{
"id": 71,
"type": "VHS_LoadVideo",
@@ -742,7 +554,7 @@
650.5508422851562
],
"flags": {},
"order": 13,
"order": 12,
"mode": 0,
"inputs": [
{
@@ -826,7 +638,7 @@
149.21942138671875
],
"flags": {},
"order": 14,
"order": 13,
"mode": 0,
"inputs": [],
"outputs": [],
@@ -909,7 +721,7 @@
"flags": {
"collapsed": true
},
"order": 15,
"order": 14,
"mode": 0,
"inputs": [],
"outputs": [
@@ -943,7 +755,7 @@
"flags": {
"collapsed": true
},
"order": 16,
"order": 15,
"mode": 2,
"inputs": [],
"outputs": [
@@ -951,7 +763,7 @@
"name": "WANVAE",
"type": "WANVAE",
"links": [
105
123
]
}
],
@@ -972,10 +784,10 @@
],
"size": [
315,
154
178
],
"flags": {},
"order": 17,
"order": 16,
"mode": 0,
"inputs": [],
"outputs": [
@@ -997,7 +809,8 @@
1,
-1,
"offload_device",
"true"
"true",
"e"
]
},
{
@@ -1012,7 +825,7 @@
226.43276977539062
],
"flags": {},
"order": 18,
"order": 17,
"mode": 0,
"inputs": [
{
@@ -1078,7 +891,7 @@
130
],
"flags": {},
"order": 19,
"order": 18,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1117,7 +930,7 @@
82
],
"flags": {},
"order": 20,
"order": 19,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1356,7 +1169,7 @@
],
"size": [
309.3501892089844,
540.900146484375
637.3502197265625
],
"flags": {},
"order": 37,
@@ -1413,13 +1226,13 @@
"hidden": false,
"paused": false,
"params": {
"filename": "control_00002.mp4",
"filename": "control_00001.mp4",
"subfolder": "",
"type": "temp",
"format": "video/h264-mp4",
"frame_rate": 16,
"workflow": "control_00002.png",
"fullpath": "N:\\AI\\ComfyUI\\temp\\control_00002.mp4"
"workflow": "control_00001.png",
"fullpath": "N:\\AI\\ComfyUI\\temp\\control_00001.mp4"
}
}
}
@@ -1438,7 +1251,7 @@
"flags": {
"collapsed": true
},
"order": 21,
"order": 20,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1569,7 +1382,7 @@
"flags": {
"collapsed": true
},
"order": 22,
"order": 21,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1678,7 +1491,7 @@
130
],
"flags": {},
"order": 23,
"order": 22,
"mode": 0,
"inputs": [],
"outputs": [
@@ -1714,7 +1527,7 @@
498.3180847167969
],
"flags": {},
"order": 24,
"order": 23,
"mode": 2,
"inputs": [],
"outputs": [
@@ -1738,134 +1551,8 @@
},
"widgets_values": [
"pasted/image (758).png",
"image"
],
"color": "#2a363b",
"bgcolor": "#3f5159"
},
{
"id": 80,
"type": "Reroute",
"pos": [
540.1338500976562,
-31.767959594726562
],
"size": [
75,
26
],
"flags": {},
"order": 43,
"mode": 0,
"inputs": [
{
"name": "",
"type": "*",
"link": 100
}
],
"outputs": [
{
"name": "",
"type": "WANVIDIMAGE_EMBEDS",
"links": []
}
],
"properties": {
"showOutputText": false,
"horizontal": false
}
},
{
"id": 63,
"type": "WanVideoImageToVideoEncode",
"pos": [
-800.4992065429688,
-261.5985107421875
],
"size": [
352.79998779296875,
338
],
"flags": {},
"order": 41,
"mode": 2,
"inputs": [
{
"name": "start_image",
"type": "IMAGE",
"link": 74
},
{
"name": "vae",
"type": "WANVAE",
"link": 105
},
{
"name": "clip_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"link": 69
},
{
"name": "end_image",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width",
"type": "INT",
"widget": {
"name": "width"
},
"link": 75
},
{
"name": "height",
"type": "INT",
"widget": {
"name": "height"
},
"link": 76
},
{
"name": "control_embeds",
"shape": 7,
"type": "WANVIDIMAGE_EMBEDS",
"link": 93
},
{
"name": "num_frames",
"type": "INT",
"widget": {
"name": "num_frames"
},
"link": 120
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_EMBEDS",
"links": [
100
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d20baf00247fd06553fdc9253e18732244e54172",
"Node name for S&R": "WanVideoImageToVideoEncode"
},
"widgets_values": [
832,
480,
81,
0.030000000000000006,
1,
1,
true,
true
"image",
""
],
"color": "#2a363b",
"bgcolor": "#3f5159"
@@ -1934,41 +1621,6 @@
},
"widgets_values": []
},
{
"id": 79,
"type": "Reroute",
"pos": [
902.3881225585938,
175.5585174560547
],
"size": [
75,
26
],
"flags": {},
"order": 42,
"mode": 0,
"inputs": [
{
"name": "",
"type": "*",
"link": 99
}
],
"outputs": [
{
"name": "",
"type": "WANVIDIMAGE_EMBEDS",
"links": [
121
]
}
],
"properties": {
"showOutputText": false,
"horizontal": false
}
},
{
"id": 78,
"type": "WanVideoControlEmbeds",
@@ -2115,7 +1767,8 @@
0,
1,
"",
"comfy"
"comfy",
""
]
},
{
@@ -2180,7 +1833,7 @@
186.99449157714844
],
"flags": {},
"order": 25,
"order": 24,
"mode": 0,
"inputs": [],
"outputs": [],
@@ -2191,6 +1844,357 @@
],
"color": "#432",
"bgcolor": "#653"
},
{
"id": 65,
"type": "WanVideoClipVisionEncode",
"pos": [
-1200.4986572265625,
-431.59814453125
],
"size": [
327.5999755859375,
262
],
"flags": {},
"order": 39,
"mode": 2,
"inputs": [
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 70
},
{
"name": "image_1",
"type": "IMAGE",
"link": 73
},
{
"name": "image_2",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "negative_image",
"shape": 7,
"type": "IMAGE",
"link": null
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"links": [
122
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d20baf00247fd06553fdc9253e18732244e54172",
"Node name for S&R": "WanVideoClipVisionEncode"
},
"widgets_values": [
1,
1,
"center",
"average",
true,
0,
0.20000000000000004
],
"color": "#233",
"bgcolor": "#355"
},
{
"id": 63,
"type": "WanVideoImageToVideoEncode",
"pos": [
-800.4992065429688,
-261.5985107421875
],
"size": [
352.79998779296875,
338
],
"flags": {},
"order": 41,
"mode": 2,
"inputs": [
{
"name": "vae",
"type": "WANVAE",
"link": 123
},
{
"name": "clip_embeds",
"type": "WANVIDIMAGE_CLIPEMBEDS",
"link": 122
},
{
"name": "start_image",
"shape": 7,
"type": "IMAGE",
"link": 124
},
{
"name": "end_image",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width",
"type": "INT",
"widget": {
"name": "width"
},
"link": 75
},
{
"name": "height",
"type": "INT",
"widget": {
"name": "height"
},
"link": 76
},
{
"name": "control_embeds",
"shape": 7,
"type": "WANVIDIMAGE_EMBEDS",
"link": 93
},
{
"name": "num_frames",
"type": "INT",
"widget": {
"name": "num_frames"
},
"link": 120
}
],
"outputs": [
{
"name": "image_embeds",
"type": "WANVIDIMAGE_EMBEDS",
"links": [
100
]
}
],
"properties": {
"cnr_id": "ComfyUI-WanVideoWrapper",
"ver": "d20baf00247fd06553fdc9253e18732244e54172",
"Node name for S&R": "WanVideoImageToVideoEncode"
},
"widgets_values": [
832,
480,
81,
0.030000000000000006,
1,
1,
true,
true
],
"color": "#2a363b",
"bgcolor": "#3f5159"
},
{
"id": 66,
"type": "ImageResizeKJ",
"pos": [
-1190.589111328125,
-70.7883529663086
],
"size": [
315,
266
],
"flags": {},
"order": 36,
"mode": 2,
"inputs": [
{
"name": "image",
"type": "IMAGE",
"link": 72
},
{
"name": "get_image_size",
"shape": 7,
"type": "IMAGE",
"link": null
},
{
"name": "width_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "width_input"
},
"link": null
},
{
"name": "height_input",
"shape": 7,
"type": "INT",
"widget": {
"name": "height_input"
},
"link": null
},
{
"name": "width",
"type": "INT",
"widget": {
"name": "width"
},
"link": 97
},
{
"name": "height",
"type": "INT",
"widget": {
"name": "height"
},
"link": 98
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
73,
124
]
},
{
"name": "width",
"type": "INT",
"links": [
75
]
},
{
"name": "height",
"type": "INT",
"links": [
76
]
}
],
"properties": {
"cnr_id": "comfyui-kjnodes",
"ver": "52c2e31a903fec2dd654fb614ea82ba2757d5028",
"Node name for S&R": "ImageResizeKJ"
},
"widgets_values": [
624,
624,
"lanczos",
false,
16,
0,
0,
"center"
]
},
{
"id": 79,
"type": "Reroute",
"pos": [
916.9335327148438,
39.80095291137695
],
"size": [
75,
26
],
"flags": {},
"order": 42,
"mode": 0,
"inputs": [
{
"name": "",
"type": "*",
"link": 99
}
],
"outputs": [
{
"name": "",
"type": "WANVIDIMAGE_EMBEDS",
"links": [
121
]
}
],
"properties": {
"showOutputText": false,
"horizontal": false
}
},
{
"id": 81,
"type": "Note",
"pos": [
562.2122802734375,
14.483749389648438
],
"size": [
255.5326690673828,
88
],
"flags": {},
"order": 25,
"mode": 0,
"inputs": [],
"outputs": [],
"properties": {},
"widgets_values": [
"Use empty embeds when doing just text to video with control, otherwise connect the image embeds from the image encoder to the sampler"
],
"color": "#432",
"bgcolor": "#653"
},
{
"id": 80,
"type": "Reroute",
"pos": [
834.6790161132812,
-45.101322174072266
],
"size": [
75,
26
],
"flags": {},
"order": 43,
"mode": 0,
"inputs": [
{
"name": "",
"type": "*",
"link": 100
}
],
"outputs": [
{
"name": "",
"type": "WANVIDIMAGE_EMBEDS",
"links": []
}
],
"properties": {
"showOutputText": false,
"horizontal": false
}
}
],
"links": [
@@ -2266,14 +2270,6 @@
6,
"TEACACHEARGS"
],
[
69,
65,
0,
63,
2,
"WANVIDIMAGE_CLIPEMBEDS"
],
[
70,
59,
@@ -2298,14 +2294,6 @@
1,
"IMAGE"
],
[
74,
66,
0,
63,
0,
"IMAGE"
],
[
75,
66,
@@ -2466,14 +2454,6 @@
0,
"WANVAE"
],
[
105,
85,
0,
63,
1,
"WANVAE"
],
[
106,
86,
@@ -2545,6 +2525,30 @@
27,
2,
"WANVIDIMAGE_EMBEDS"
],
[
122,
65,
0,
63,
1,
"WANVIDIMAGE_CLIPEMBEDS"
],
[
123,
85,
0,
63,
0,
"WANVAE"
],
[
124,
66,
0,
63,
2,
"IMAGE"
]
],
"groups": [
@@ -2591,10 +2595,10 @@
"config": {},
"extra": {
"ds": {
"scale": 0.8264462809917354,
"scale": 1.1,
"offset": [
324.06772332943353,
996.9959463677691
177.23656305995507,
361.5261810657517
]
},
"node_versions": {
nodes.py
+33 -20
View File
@@ -1352,7 +1352,6 @@ class WanVideoImageToVideoEncode:
@classmethod
def INPUT_TYPES(s):
return {"required": {
"start_image": ("IMAGE", {"tooltip": "Image to encode"}),
"vae": ("WANVAE",),
"width": ("INT", {"default": 832, "min": 64, "max": 2048, "step": 8, "tooltip": "Width of the image to encode"}),
"height": ("INT", {"default": 480, "min": 64, "max": 29048, "step": 8, "tooltip": "Height of the image to encode"}),
@@ -1364,6 +1363,7 @@ class WanVideoImageToVideoEncode:
"force_offload": ("BOOLEAN", {"default": True}),
},
"optional": {
"start_image": ("IMAGE", {"tooltip": "Image to encode"}),
"end_image": ("IMAGE", {"tooltip": "end frame"}),
"control_embeds": ("WANVIDIMAGE_EMBEDS", {"tooltip": "Control signal for the Fun -model"}),
"fun_model": ("BOOLEAN", {"default": False, "tooltip": "Enable when using Fun model"}),
@@ -1375,8 +1375,8 @@ class WanVideoImageToVideoEncode:
FUNCTION = "process"
CATEGORY = "WanVideoWrapper"
def process(self, vae, start_image, width, height, num_frames, clip_embeds, force_offload, noise_aug_strength,
start_latent_strength, end_latent_strength, end_image=None, control_embeds=None, fun_model=False):
def process(self, vae, width, height, num_frames, clip_embeds, force_offload, noise_aug_strength,
start_latent_strength, end_latent_strength, start_image=None, end_image=None, control_embeds=None, fun_model=False):
device = mm.get_torch_device()
offload_device = mm.unet_offload_device()
@@ -1390,11 +1390,13 @@ class WanVideoImageToVideoEncode:
lat_w = W // 8
num_frames = ((num_frames - 1) // 4) * 4 + 1
two_ref_images = start_image is not None and end_image is not None
base_frames = num_frames + (1 if end_image is not None and not fun_model else 0)
base_frames = num_frames + (1 if two_ref_images and not fun_model else 0)
mask = torch.zeros(1, base_frames, lat_h, lat_w, device=device)
mask[:, 0] = 1 # First frame
if start_image is not None:
mask[:, 0] = 1 # First frame
if end_image is not None and not fun_model:
mask[:, -1] = 1 # End frame if exists
@@ -1411,10 +1413,11 @@ class WanVideoImageToVideoEncode:
mask = mask.movedim(1, 2)[0]# C, T, H, W
# Resize and rearrange the input image dimensions
resized_image = common_upscale(start_image.movedim(-1, 1), W, H, "lanczos", "disabled").movedim(0, 1)
resized_image = resized_image * 2 - 1
if noise_aug_strength > 0.0:
resized_image = add_noise_to_reference_video(resized_image, ratio=noise_aug_strength)
if start_image is not None:
resized_start_image = common_upscale(start_image.movedim(-1, 1), W, H, "lanczos", "disabled").movedim(0, 1)
resized_start_image = resized_start_image * 2 - 1
if noise_aug_strength > 0.0:
resized_start_image = add_noise_to_reference_video(resized_start_image, ratio=noise_aug_strength)
if end_image is not None:
resized_end_image = common_upscale(end_image.movedim(-1, 1), W, H, "lanczos", "disabled").movedim(0, 1)
@@ -1425,23 +1428,31 @@ class WanVideoImageToVideoEncode:
# Concatenate image with zero frames and encode
vae.to(device)
if end_image is None:
if start_image is not None and end_image is None:
zero_frames = torch.zeros(3, num_frames-1, H, W, device=device)
concatenated = torch.cat([resized_image.to(device), zero_frames], dim=1)
concatenated = torch.cat([resized_start_image.to(device), zero_frames], dim=1)
elif start_image is None and end_image is not None:
zero_frames = torch.zeros(3, num_frames-1, H, W, device=device)
concatenated = torch.cat([zero_frames, resized_end_image.to(device)], dim=1)
elif start_image is None and end_image is None:
concatenated = torch.zeros(3, num_frames, H, W, device=device)
else:
if fun_model:
zero_frames = torch.zeros(3, num_frames-2, H, W, device=device)
else:
zero_frames = torch.zeros(3, num_frames-1, H, W, device=device)
concatenated = torch.cat([resized_image.to(device) * start_latent_strength, zero_frames, resized_end_image.to(device) * end_latent_strength ], dim=1)
concatenated = torch.cat([resized_start_image.to(device), zero_frames, resized_end_image.to(device)], dim=1)
y = vae.encode([concatenated.to(device=device, dtype=vae.dtype)], device, end_=(end_image is not None and not fun_model))[0]
y[:, :1] *= start_latent_strength
y[:, -1:] *= end_latent_strength
if control_embeds is None:
y = torch.cat([mask, y])
else:
if end_image is None:
y[:, 1:] = 0
elif start_image is None:
y[:, -1:] = 0
else:
y[:, 1:-1] = 0 # doesn't seem to work anyway though...
@@ -1538,12 +1549,14 @@ class WanVideoControlEmbeds:
"max_seq_len": seq_len,
"target_shape": samples.shape,
"num_frames": num_frames,
"control_images": samples,
"start_percent": start_percent,
"end_percent": end_percent,
"control_embeds": {
"control_images": samples,
"start_percent": start_percent,
"end_percent": end_percent
}
}
return ({"control_embeds": embeds},)
return (embeds,)
class WanVideoSLG:
@classmethod
@@ -2088,7 +2101,6 @@ class WanVideoSampler:
image_cond_input = torch.cat([torch.zeros_like(image_cond), image_cond])
if control_lora:
image_cond_input = None
if not control_start_percent <= current_step_percentage <= control_end_percent:
control_lora_enabled = False
if patcher.model.is_patched:
@@ -2096,6 +2108,7 @@ class WanVideoSampler:
patcher.unpatch_model(device)
patcher.model.is_patched = False
else:
image_cond_input = control_latents.to(device)
if not patcher.model.is_patched:
log.info("Loading LoRA...")
patcher = apply_lora(patcher, device, device, low_mem_load=False)
@@ -2109,7 +2122,7 @@ class WanVideoSampler:
'freqs': freqs,
't': timestep,
'current_step': idx,
'y': [image_cond_input] if image_cond is not None else None,
'y': [image_cond_input] if image_cond_input is not None else None,
'control_lora_enabled': control_lora_enabled,
}