ComfyPilotComfyPilot
工作流 / 音频 / MelBandRoFormer音频分离
MelBandRoFormer音频分离

MelBandRoFormer音频分离

MelBandRoFormer Audio Separation

加载音频文件并使用MelBandRoFormer模型将其分离为人声和伴奏音轨。输出两个MP3文件，分别包含独立人声和背景音乐。

音频 6 个节点 1 个输入 35 次使用 ComfyUI ≥ 0.25.0 ComfyUI 5 KB

下载工作流 JSON 查看官方来源

下载后把文件拖进 ComfyUI 窗口即可打开。

需要的权重

这个工作流引用了 1 个权重文件，其中 1 个能在本站模型库里找到。

MelBandRoformer_fp16.safetensors 在模型库查看

节点构成

SaveAudioMP3 × 2
LoadAudio × 1
MelBandRoFormerModelLoader × 1
MelBandRoFormerSampler × 1
MarkdownNote × 1

官方标注的模型

None

工作流 JSON

和上面「下载工作流 JSON」是同一份文件。

1{
2 "id": "21922001-4ca7-45b7-84dc-56fcfe7d44a9",
3 "revision": 0,
4 "last_node_id": 7,
5 "last_link_id": 4,
6 "nodes": [
7 {
8 "id": 1,
9 "type": "LoadAudio",
10 "pos": [
11 120,
12 430
13 ],
14 "size": [
15 320,
16 176
17 ],
18 "flags": [],
19 "order": 0,
20 "mode": 0,
21 "inputs": [],
22 "outputs": [
23 {
24 "name": "AUDIO",
25 "type": "AUDIO",
26 "links": [
27 1
28 ]
29 }
30 ],
31 "properties": {
32 "Node name for S&R": "LoadAudio"
33 },
34 "widgets_values": [
35 "melbandroformer_input.mp3",
36 null,
37 null
38 ]
39 },
40 {
41 "id": 2,
42 "type": "MelBandRoFormerModelLoader",
43 "pos": [
44 120,
45 200
46 ],
47 "size": [
48 320,
49 170
50 ],
51 "flags": [],
52 "order": 1,
53 "mode": 0,
54 "inputs": [],
55 "outputs": [
56 {
57 "name": "model",
58 "type": "MELROFORMERMODEL",
59 "links": [
60 2
61 ]
62 }
63 ],
64 "title": "MelBand Model Loader",
65 "properties": {
66 "Node name for S&R": "MelBandRoFormerModelLoader"
67 },
68 "widgets_values": [
69 "MelBandRoformer_fp16.safetensors"
70 ]
71 },
72 {
73 "id": 3,
74 "type": "MelBandRoFormerSampler",
75 "pos": [
76 500.000010648165,
77 200.00000116712488
78 ],
79 "size": [
80 320,
81 170
82 ],
83 "flags": [],
84 "order": 3,
85 "mode": 0,
86 "inputs": [
87 {
88 "name": "model",
89 "type": "MELROFORMERMODEL",
90 "link": 2
91 },
92 {
93 "name": "audio",
94 "type": "AUDIO",
95 "link": 1
96 }
97 ],
98 "outputs": [
99 {
100 "name": "vocals",
101 "type": "AUDIO",
102 "links": [
103 3
104 ]
105 },
106 {
107 "name": "instruments",
108 "type": "AUDIO",
109 "links": [
110 4
111 ]
112 }
113 ],
114 "title": "MelBand Separator",
115 "properties": {
116 "Node name for S&R": "MelBandRoFormerSampler"
117 },
118 "widgets_values": []
119 },
120 {
121 "id": 4,
122 "type": "SaveAudioMP3",
123 "pos": [
124 900,
125 200
126 ],
127 "size": [
128 320,
129 210.65625
130 ],
131 "flags": [],
132 "order": 4,
133 "mode": 0,
134 "inputs": [
135 {
136 "name": "audio",
137 "type": "AUDIO",
138 "link": 3
139 }
140 ],
141 "outputs": [],
142 "title": "Save Vocals",
143 "properties": {
144 "Node name for S&R": "SaveAudioMP3"
145 },
146 "widgets_values": [
147 "vocals",
148 "128k"
149 ]
150 },
151 {
152 "id": 5,
153 "type": "SaveAudioMP3",
154 "pos": [
155 900,
156 460
157 ],
158 "size": [
159 320,
160 210.65625
161 ],
162 "flags": [],
163 "order": 5,
164 "mode": 0,
165 "inputs": [
166 {
167 "name": "audio",
168 "type": "AUDIO",
169 "link": 4
170 }
171 ],
172 "outputs": [],
173 "title": "Save Instruments",
174 "properties": {
175 "Node name for S&R": "SaveAudioMP3"
176 },
177 "widgets_values": [
178 "instruments",
179 "128k"
180 ]
181 },
182 {
183 "id": 7,
184 "type": "MarkdownNote",
185 "pos": [
186 -430,
187 200
188 ],
189 "size": [
190 480,
191 596.46875
192 ],
193 "flags": [],
194 "order": 2,
195 "mode": 0,
196 "inputs": [],
197 "outputs": [],
198 "properties": [],
199 "widgets_values": [
200 "# MelBandRoFormer Audio Separation\n\nSeparates any audio track into vocals and instruments using the MelBandRoFormer model.\n\n## Workflow\n\n1. **LoadAudio** — loads your input audio file\n2. **MelBandRoFormerModelLoader** — loads `MelBandRoformer_fp16.safetensors` from `models/diffusion_models/`\n3. **MelBandRoFormerSampler** — processes audio and outputs two stems: vocals and instruments\n4. **SaveAudioMP3** × 2 — saves each stem as a separate MP3\n## Requirements\n\n- **Node:** [ComfyUI-MelBandRoFormer](https://github.com/kijai/ComfyUI-MelBandRoFormer) \n- **Model:** [MelBandRoformer_fp16.safetensors](https://huggingface.co/Kijai/MelBandRoFormer_comfy/resolve/main/MelBandRoformer_fp16.safetensors) (435.3 MB) (~456 MB) Put it under `ComfyUI/models/diffusion_models`\n\n> Input audio should be stereo for best results."
201 ],
202 "color": "#222",
203 "bgcolor": "#000"
204 }
205 ],
206 "links": [
207 [
208 1,
209 1,
210 0,
211 3,
212 1,
213 "AUDIO"
214 ],
215 [
216 2,
217 2,
218 0,
219 3,
220 0,
221 "MELROFORMERMODEL"
222 ],
223 [
224 3,
225 3,
226 0,
227 4,
228 0,
229 "AUDIO"
230 ],
231 [
232 4,
233 3,
234 1,
235 5,
236 0,
237 "AUDIO"
238 ]
239 ],
240 "groups": [],
241 "config": [],
242 "extra": {
243 "ds": {
244 "scale": 0.6209213230591553,
245 "offset": [
246 935.4804215346629,
247 377.78710221657855
248 ]
249 },
250 "workflowRendererVersion": "Vue-corrected",
251 "frontendVersion": "1.41.13",
252 "VHS_latentpreview": false,
253 "VHS_latentpreviewrate": 0,
254 "VHS_MetadataImage": true,
255 "VHS_KeepIntermediate": true
256 },
257 "version": 0.4
258}