工作流 /
视频 /
Higgsfield Genjutsu：动作迁移
Higgsfield Genjutsu：动作迁移
Higgsfield Genjutsu: Motion Transfer
使用 Higgsfield Genjutsu 重铸现有视频：来源 CLIP 保留其运动、相机移动、时序和音频，而演员、场景和风格则根据 1 到 8 张参考图像重建。在可选提示词中描述哪张参考图像替换什么内容。输出长度跟随来源 CLIP。非常适合重新设计素材风格、替换镜头中的演员，以及构建拍摄从未覆盖的场景。
视频
6 个节点 3 个输入 215 次使用 ComfyUI ≥ 0.40.0 ComfyUI 8 KB
节点构成
LoadImage
× 2
HiggsfieldMotionTransferNode
× 1
SaveVideo
× 1
MarkdownNote
× 1
LoadVideo
× 1
官方标注的模型
Genjutsu 1.0
工作流 JSON
和上面「下载工作流 JSON」是同一份文件。
1{
2 "id": "8f2c1d64-0f5b-4a17-9d3e-5b7c9a1e4d22",
3 "revision": 0,
4 "last_node_id": 15,
5 "last_link_id": 7,
6 "nodes": [
7 {
8 "id": 9,
9 "type": "HiggsfieldMotionTransferNode",
10 "pos": [
11 -412,
12 120
13 ],
14 "size": [
15 480,
16 490
17 ],
18 "flags": [],
19 "order": 4,
20 "mode": 0,
21 "inputs": [
22 {
23 "name": "model.video",
24 "type": "VIDEO",
25 "link": 5
26 },
27 {
28 "name": "model.reference_images.image_1",
29 "type": "IMAGE",
30 "link": 6
31 },
32 {
33 "name": "model.reference_images.image_2",
34 "shape": 7,
35 "type": "IMAGE",
36 "link": 7
37 },
38 {
39 "name": "model.reference_images.image_3",
40 "shape": 7,
41 "type": "IMAGE",
42 "link": null
43 }
44 ],
45 "outputs": [
46 {
47 "name": "VIDEO",
48 "type": "VIDEO",
49 "links": [
50 4
51 ]
52 }
53 ],
54 "properties": {
55 "Node name for S&R": "HiggsfieldMotionTransferNode"
56 },
57 "widgets_values": [
58 "genjutsu-v1.0",
59 "Recast the dancer as the snow queen from the first reference image, and use the second reference image for the new setting. Keep the source motion, camera movement, timing and audio.",
60 "720p",
61 42,
62 "randomize",
63 "randomize"
64 ],
65 "widgets_values_named": {
66 "model": "genjutsu-v1.0",
67 "model.prompt": "Recast the dancer as the snow queen from the first reference image, and use the second reference image for the new setting. Keep the source motion, camera movement, timing and audio.",
68 "model.resolution": "720p",
69 "model.seed": 42,
70 "control_after_generate#1": "randomize",
71 "control_after_generate": "randomize"
72 },
73 "color": "#432",
74 "bgcolor": "#653"
75 },
76 {
77 "id": 10,
78 "type": "SaveVideo",
79 "pos": [
80 116,
81 120
82 ],
83 "size": [
84 440,
85 460
86 ],
87 "flags": [],
88 "order": 5,
89 "mode": 0,
90 "inputs": [
91 {
92 "name": "video",
93 "type": "VIDEO",
94 "link": 4
95 }
96 ],
97 "outputs": [
98 {
99 "name": "video",
100 "type": "VIDEO",
101 "links": null
102 }
103 ],
104 "properties": {
105 "Node name for S&R": "SaveVideo"
106 },
107 "widgets_values": [
108 "video/genjutsu_motion_transfer",
109 "auto",
110 "auto",
111 "auto"
112 ],
113 "widgets_values_named": {
114 "filename_prefix": "video/genjutsu_motion_transfer",
115 "format": "auto",
116 "format.codec": "auto",
117 "codec": "auto"
118 }
119 },
120 {
121 "id": 12,
122 "type": "MarkdownNote",
123 "pos": [
124 -1500,
125 120
126 ],
127 "size": [
128 620,
129 700
130 ],
131 "flags": [],
132 "order": 0,
133 "mode": 0,
134 "inputs": [],
135 "outputs": [],
136 "title": "Higgsfield Genjutsu Motion Transfer",
137 "properties": [],
138 "widgets_values": [
139 "## Higgsfield Genjutsu Motion Transfer\n\nRecast a video: the source clip keeps its motion, camera movement, timing and audio, while the cast, location and style are rebuilt from 1 to 8 reference images.\n\nThis example turns a rooftop dancer into a snow queen, first with a character reference, then with an optional scene reference.\n\n## Input Assets\n\n- [rooftop_dancer.mp4](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/rooftop_dancer.mp4)\n- [snow_queen_ref.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/snow_queen_ref.png)\n- [ice_palace_ref.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/ice_palace_ref.png)\n\n**Inputs:**\n\n- `video`: source motion clip. At least 4 seconds, only the first 30 seconds are used.\n- `reference_images`: 1 to 8 images of the characters, locations or styles to use. Click **+** on the input to add more references.\n- `prompt`: optional. Describe which reference replaces what, for example which person it should replace.\n- `resolution`: 480p, 720p (default) or 1080p.\n\n**Notes:**\n\n- Output length follows the source clip, and the model is billed per output second.\n- One partner job runs at a time.\n- The seed only controls whether the node re-runs: results are non-deterministic either way.\n"
140 ],
141 "widgets_values_named": {
142 "text": "## Higgsfield Genjutsu Motion Transfer\n\nRecast a video: the source clip keeps its motion, camera movement, timing and audio, while the cast, location and style are rebuilt from 1 to 8 reference images.\n\nThis example turns a rooftop dancer into a snow queen, first with a character reference, then with an optional scene reference.\n\n## Input Assets\n\n- [rooftop_dancer.mp4](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/rooftop_dancer.mp4)\n- [snow_queen_ref.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/snow_queen_ref.png)\n- [ice_palace_ref.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/ice_palace_ref.png)\n\n**Inputs:**\n\n- `video`: source motion clip. At least 4 seconds, only the first 30 seconds are used.\n- `reference_images`: 1 to 8 images of the characters, locations or styles to use. Click **+** on the input to add more references.\n- `prompt`: optional. Describe which reference replaces what, for example which person it should replace.\n- `resolution`: 480p, 720p (default) or 1080p.\n\n**Notes:**\n\n- Output length follows the source clip, and the model is billed per output second.\n- One partner job runs at a time.\n- The seed only controls whether the node re-runs: results are non-deterministic either way.\n"
143 },
144 "color": "#222",
145 "bgcolor": "#000"
146 },
147 {
148 "id": 13,
149 "type": "LoadVideo",
150 "pos": [
151 -800,
152 120
153 ],
154 "size": [
155 340,
156 460
157 ],
158 "flags": [],
159 "order": 1,
160 "mode": 0,
161 "inputs": [],
162 "outputs": [
163 {
164 "name": "VIDEO",
165 "type": "VIDEO",
166 "links": [
167 5
168 ]
169 }
170 ],
171 "properties": {
172 "Node name for S&R": "LoadVideo"
173 },
174 "widgets_values": [
175 "rooftop_dancer.mp4",
176 "image"
177 ],
178 "widgets_values_named": {
179 "file": "rooftop_dancer.mp4",
180 "upload": "image"
181 }
182 },
183 {
184 "id": 14,
185 "type": "LoadImage",
186 "pos": [
187 -790,
188 670
189 ],
190 "size": [
191 330,
192 420
193 ],
194 "flags": [],
195 "order": 2,
196 "mode": 0,
197 "inputs": [],
198 "outputs": [
199 {
200 "name": "IMAGE",
201 "type": "IMAGE",
202 "links": [
203 6
204 ]
205 },
206 {
207 "name": "MASK",
208 "type": "MASK",
209 "links": null
210 }
211 ],
212 "properties": {
213 "Node name for S&R": "LoadImage"
214 },
215 "widgets_values": [
216 "snow_queen_ref.png",
217 "image"
218 ],
219 "widgets_values_named": {
220 "image": "snow_queen_ref.png",
221 "upload": "image"
222 }
223 },
224 {
225 "id": 15,
226 "type": "LoadImage",
227 "pos": [
228 -780,
229 1170
230 ],
231 "size": [
232 320,
233 400
234 ],
235 "flags": [],
236 "order": 3,
237 "mode": 0,
238 "inputs": [],
239 "outputs": [
240 {
241 "name": "IMAGE",
242 "type": "IMAGE",
243 "links": [
244 7
245 ]
246 },
247 {
248 "name": "MASK",
249 "type": "MASK",
250 "links": null
251 }
252 ],
253 "properties": {
254 "Node name for S&R": "LoadImage"
255 },
256 "widgets_values": [
257 "ice_palace_ref.png",
258 "image"
259 ],
260 "widgets_values_named": {
261 "image": "ice_palace_ref.png",
262 "upload": "image"
263 }
264 }
265 ],
266 "links": [
267 [
268 4,
269 9,
270 0,
271 10,
272 0,
273 "VIDEO"
274 ],
275 [
276 5,
277 13,
278 0,
279 9,
280 0,
281 "VIDEO"
282 ],
283 [
284 6,
285 14,
286 0,
287 9,
288 1,
289 "IMAGE"
290 ],
291 [
292 7,
293 15,
294 0,
295 9,
296 2,
297 "IMAGE"
298 ]
299 ],
300 "groups": [],
301 "config": [],
302 "extra": {
303 "ds": {
304 "scale": 0.7556985851288385,
305 "offset": [
306 1652.9500003487594,
307 -43.85404560880734
308 ]
309 },
310 "frontendVersion": "1.55.14"
311 },
312 "version": 0.4
313}