工作流 /
图像工具 /
图像放大：Topaz Wonder 4
图像放大：Topaz Wonder 4
Image Upscale: Topaz Wonder 4
使用 Topaz 的 Wonder 4 生成模型放大柔和、压缩的图像：它会重建合理的细节，而不是拉伸像素，在 1x 到 6x 的放大倍率下保持撕裂边缘、接缝和纯色区域完好。该工作流接收 1 张输入图像，并在生成 1 个增强输出的同时提供并排对比。非常适合需要从较小或压缩的来源中保留质量的拼贴、插画和图形作品。
图像工具
7 个节点 1 个输入 96 次使用 ComfyUI ≥ 0.44.0 ComfyUI 11 KB
节点构成
LoadImage
× 1
TopazImageEnhanceV2
× 1
ImageScaleBy
× 1
GetImageSize
× 1
SaveImageAdvanced
× 1
ImageCompare
× 1
MarkdownNote
× 1
官方标注的模型
Topaz
工作流 JSON
和上面「下载工作流 JSON」是同一份文件。
1{
2 "id": "cda934e1-5fac-49e9-8da3-73e91c3c41ec",
3 "revision": 0,
4 "last_node_id": 7,
5 "last_link_id": 8,
6 "nodes": [
7 {
8 "id": 1,
9 "type": "LoadImage",
10 "pos": [
11 -974.0016132778537,
12 1566.1140654894616
13 ],
14 "size": [
15 410.7603866832164,
16 526.175903695124
17 ],
18 "flags": [],
19 "order": 0,
20 "mode": 0,
21 "inputs": [],
22 "outputs": [
23 {
24 "name": "IMAGE",
25 "type": "IMAGE",
26 "links": [
27 1,
28 4,
29 7
30 ]
31 },
32 {
33 "name": "MASK",
34 "type": "MASK",
35 "links": []
36 }
37 ],
38 "properties": {
39 "Node name for S&R": "LoadImage"
40 },
41 "widgets_values": [
42 "topaz_wonder4_collage_lowres.png",
43 "image"
44 ],
45 "widgets_values_named": {
46 "image": "topaz_wonder4_collage_lowres.png",
47 "upload": "image"
48 }
49 },
50 {
51 "id": 2,
52 "type": "TopazImageEnhanceV2",
53 "pos": [
54 -515.2412265946373,
55 1566.1140654894616
56 ],
57 "size": [
58 400,
59 106
60 ],
61 "flags": [],
62 "order": 4,
63 "mode": 0,
64 "showAdvanced": true,
65 "inputs": [
66 {
67 "name": "image",
68 "type": "IMAGE",
69 "link": 1
70 },
71 {
72 "name": "output_width",
73 "shape": 7,
74 "type": "INT",
75 "widget": {
76 "name": "output_width"
77 },
78 "link": 2
79 },
80 {
81 "name": "output_height",
82 "shape": 7,
83 "type": "INT",
84 "widget": {
85 "name": "output_height"
86 },
87 "link": 3
88 }
89 ],
90 "outputs": [
91 {
92 "name": "IMAGE",
93 "type": "IMAGE",
94 "links": [
95 6
96 ]
97 }
98 ],
99 "properties": {
100 "Node name for S&R": "TopazImageEnhanceV2"
101 },
102 "widgets_values": [
103 "Wonder 4",
104 0,
105 0
106 ],
107 "widgets_values_named": {
108 "model": "Wonder 4",
109 "output_width": 0,
110 "output_height": 0
111 },
112 "color": "#432",
113 "bgcolor": "#653"
114 },
115 {
116 "id": 3,
117 "type": "ImageScaleBy",
118 "pos": [
119 -461.46380111337453,
120 1975.1225707586636
121 ],
122 "size": [
123 270,
124 82
125 ],
126 "flags": [],
127 "order": 2,
128 "mode": 0,
129 "inputs": [
130 {
131 "name": "image",
132 "type": "IMAGE",
133 "link": 4
134 }
135 ],
136 "outputs": [
137 {
138 "name": "IMAGE",
139 "type": "IMAGE",
140 "links": [
141 5
142 ]
143 }
144 ],
145 "properties": {
146 "Node name for S&R": "ImageScaleBy"
147 },
148 "widgets_values": [
149 "nearest-exact",
150 2
151 ],
152 "widgets_values_named": {
153 "upscale_method": "nearest-exact",
154 "scale_by": 2
155 },
156 "color": "#322",
157 "bgcolor": "#533"
158 },
159 {
160 "id": 4,
161 "type": "GetImageSize",
162 "pos": [
163 -463.87373489260676,
164 1848.7056716977181
165 ],
166 "size": [
167 268.96622216789973,
168 73.97565349916178
169 ],
170 "flags": [],
171 "order": 3,
172 "mode": 0,
173 "inputs": [
174 {
175 "name": "image",
176 "type": "IMAGE",
177 "link": 5
178 }
179 ],
180 "outputs": [
181 {
182 "name": "width",
183 "type": "INT",
184 "links": [
185 2
186 ]
187 },
188 {
189 "name": "height",
190 "type": "INT",
191 "links": [
192 3
193 ]
194 },
195 {
196 "name": "batch_size",
197 "type": "INT",
198 "links": []
199 }
200 ],
201 "properties": {
202 "Node name for S&R": "GetImageSize"
203 }
204 },
205 {
206 "id": 5,
207 "type": "SaveImageAdvanced",
208 "pos": [
209 -67.24122659463728,
210 1566.1140654894616
211 ],
212 "size": [
213 503.13990351753694,
214 498.23242973804986
215 ],
216 "flags": [],
217 "order": 5,
218 "mode": 0,
219 "inputs": [
220 {
221 "name": "images",
222 "type": "IMAGE",
223 "link": 6
224 }
225 ],
226 "outputs": [
227 {
228 "name": "images",
229 "type": "IMAGE",
230 "links": [
231 8
232 ]
233 }
234 ],
235 "properties": [],
236 "widgets_values": [
237 "api_topaz_image_enhance_wonder4",
238 "png",
239 "8-bit",
240 "sRGB"
241 ],
242 "widgets_values_named": {
243 "filename_prefix": "api_topaz_image_enhance_wonder4",
244 "format": "png",
245 "format.bit_depth": "8-bit",
246 "format.input_color_space": "sRGB"
247 }
248 },
249 {
250 "id": 6,
251 "type": "ImageCompare",
252 "pos": [
253 483.89867692289954,
254 1566.1140654894616
255 ],
256 "size": [
257 469.66320080407945,
258 506.66329967565093
259 ],
260 "flags": [],
261 "order": 6,
262 "mode": 0,
263 "inputs": [
264 {
265 "name": "image_a",
266 "shape": 7,
267 "type": "IMAGE",
268 "link": 7
269 },
270 {
271 "name": "image_b",
272 "shape": 7,
273 "type": "IMAGE",
274 "link": 8
275 }
276 ],
277 "outputs": [],
278 "properties": {
279 "Node name for S&R": "ImageCompare"
280 },
281 "widgets_values": [
282 [
283 "",
284 ""
285 ]
286 ],
287 "widgets_values_named": {
288 "compare_view": [
289 "",
290 ""
291 ]
292 }
293 },
294 {
295 "id": 7,
296 "type": "MarkdownNote",
297 "pos": [
298 -1679.0103789273157,
299 1561.1460386314257
300 ],
301 "size": [
302 620.9812524354631,
303 629.6076024695928
304 ],
305 "flags": [],
306 "order": 1,
307 "mode": 0,
308 "inputs": [],
309 "outputs": [],
310 "properties": [],
311 "widgets_values": [
312 "## Topaz Wonder 4: Image Upscale\n\n**Bring a small, soft image back.** Wonder 4 rebuilds plausible detail instead of stretching pixels, so a soft, lightly compressed copy returns clean at 2 to 6 times its size.\n\n---\n\n### What this graph does\n\n1. **Load Image**: the small, soft source. The bundled demo is a 512 x 512, lightly compressed copy of a 1024 x 1024 paper-collage image.\n2. **Image Scale By**: sets the target. `scale_by` 2 means twice the input, so the demo comes back at the original 1024 x 1024.\n3. **Image Size**: reads that scaled size and passes it to the node as `output_width` and `output_height`.\n4. **Topaz Image Enhance** with `model` set to `Wonder 4`: upscales to the target size.\n5. **Save Image (Advanced)** writes the result, **Compare Images** shows the soft input next to it.\n\n---\n\n### Controls\n\n| Control | What it does |\n|---|---|\n| `model` | `Wonder 4` is a single pass at the target size. `Wonder 3.5` and `Bloom 2` add their own settings, such as Low / Medium / High enhancement or a prompt |\n| `scale_by` | Upscale factor: 1x to 6x on Wonder 4 and Wonder 3.5, up to 12x on Bloom 2 |\n| `output_width` / `output_height` | Filled here from Image Size. Zero means automatic |\n| `filename_prefix` | Where the result lands in the output folder |\n\nWonder keeps the source aspect ratio: when the requested width and height disagree, the larger scale wins and the output is bigger than asked. The node accepts up to 179 megapixels and 32000 pixels per side, and it downscales the input to about 16 megapixels before sending it.\n\n### When Wonder 4 fits\n\nBuilt for small, medium, soft, noisy or compressed images: old scans, screenshots, AI thumbnails, phone crops. A graphic, textured image is a good test, since torn edges, seams and flat color fields show at once whether the model invents detail or keeps the original. For an already sharp, high resolution photo a precision upscaler keeps more of the original.\n"
313 ],
314 "widgets_values_named": {
315 "text": "## Topaz Wonder 4: Image Upscale\n\n**Bring a small, soft image back.** Wonder 4 rebuilds plausible detail instead of stretching pixels, so a soft, lightly compressed copy returns clean at 2 to 6 times its size.\n\n---\n\n### What this graph does\n\n1. **Load Image**: the small, soft source. The bundled demo is a 512 x 512, lightly compressed copy of a 1024 x 1024 paper-collage image.\n2. **Image Scale By**: sets the target. `scale_by` 2 means twice the input, so the demo comes back at the original 1024 x 1024.\n3. **Image Size**: reads that scaled size and passes it to the node as `output_width` and `output_height`.\n4. **Topaz Image Enhance** with `model` set to `Wonder 4`: upscales to the target size.\n5. **Save Image (Advanced)** writes the result, **Compare Images** shows the soft input next to it.\n\n---\n\n### Controls\n\n| Control | What it does |\n|---|---|\n| `model` | `Wonder 4` is a single pass at the target size. `Wonder 3.5` and `Bloom 2` add their own settings, such as Low / Medium / High enhancement or a prompt |\n| `scale_by` | Upscale factor: 1x to 6x on Wonder 4 and Wonder 3.5, up to 12x on Bloom 2 |\n| `output_width` / `output_height` | Filled here from Image Size. Zero means automatic |\n| `filename_prefix` | Where the result lands in the output folder |\n\nWonder keeps the source aspect ratio: when the requested width and height disagree, the larger scale wins and the output is bigger than asked. The node accepts up to 179 megapixels and 32000 pixels per side, and it downscales the input to about 16 megapixels before sending it.\n\n### When Wonder 4 fits\n\nBuilt for small, medium, soft, noisy or compressed images: old scans, screenshots, AI thumbnails, phone crops. A graphic, textured image is a good test, since torn edges, seams and flat color fields show at once whether the model invents detail or keeps the original. For an already sharp, high resolution photo a precision upscaler keeps more of the original.\n"
316 },
317 "color": "#222",
318 "bgcolor": "#000"
319 }
320 ],
321 "links": [
322 [
323 4,
324 1,
325 0,
326 3,
327 0,
328 "IMAGE"
329 ],
330 [
331 5,
332 3,
333 0,
334 4,
335 0,
336 "IMAGE"
337 ],
338 [
339 7,
340 1,
341 0,
342 6,
343 0,
344 "IMAGE"
345 ],
346 [
347 8,
348 5,
349 0,
350 6,
351 1,
352 "IMAGE"
353 ],
354 [
355 1,
356 1,
357 0,
358 2,
359 0,
360 "IMAGE"
361 ],
362 [
363 2,
364 4,
365 0,
366 2,
367 1,
368 "INT"
369 ],
370 [
371 3,
372 4,
373 1,
374 2,
375 2,
376 "INT"
377 ],
378 [
379 6,
380 2,
381 0,
382 5,
383 0,
384 "IMAGE"
385 ]
386 ],
387 "groups": [],
388 "config": [],
389 "extra": {
390 "ds": {
391 "scale": 0.48616908261366126,
392 "offset": [
393 1789.4267317195702,
394 -968.2895784730329
395 ]
396 },
397 "frontendVersion": "1.57.6"
398 },
399 "version": 0.4
400}