工作流 / 音频 / VEED Clean Audio：语音降噪

VEED Clean Audio：语音降噪

VEED Clean Audio: Speech Cleanup

用 VEED Clean Audio 去掉语音录音里的底噪：音乐与音效会随噪声一起被移除，句间停顿被彻底清理，输出 48 kHz 单声道音频，并可按目标响度归一化。单段音频最长 1799 秒。

音频 4 个节点 1 个输入 21 次使用 ComfyUI ≥ 0.43.0 ComfyUI 4 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。

节点构成

VeedCleanAudioNode × 1

LoadAudio × 1

SaveAudioAdvanced × 1

MarkdownNote × 1

官方标注的模型

VEED

工作流 JSON

和上面「下载工作流 JSON」是同一份文件。

1{
  "id": "5b4ffc84-79bf-4ead-840c-689638e237ab",
  "revision": 0,
  "last_node_id": 15,
  "last_link_id": 8,
  "nodes": [
      {
          "id": 12,
          "type": "VeedCleanAudioNode",
          "pos": [
              2168,
              1010
          ],
          "size": [
              330,
              178
          ],
          "flags": [],
          "order": 2,
          "mode": 0,
          "inputs": [
              {
                  "name": "audio",
                  "type": "AUDIO",
                  "link": 7
              }
          ],
          "outputs": [
              {
                  "name": "AUDIO",
                  "type": "AUDIO",
                  "links": [
                      8
                  ]
              }
          ],
          "properties": {
              "Node name for S&R": "VeedCleanAudioNode"
          },
          "widgets_values": [
              0.874,
              "normalize",
              -19
          ],
          "widgets_values_named": {
              "strength": 0.874,
              "loudness": "normalize",
              "loudness.target_lufs": -19
          },
          "color": "#432",
          "bgcolor": "#653"
      },
      {
          "id": 13,
          "type": "LoadAudio",
          "pos": [
              1730,
              1010
          ],
          "size": [
              390,
              121.94446408248871
          ],
          "flags": [],
          "order": 0,
          "mode": 0,
          "inputs": [],
          "outputs": [
              {
                  "name": "AUDIO",
                  "type": "AUDIO",
                  "links": [
                      7
                  ]
              }
          ],
          "properties": {
              "Node name for S&R": "LoadAudio"
          },
          "widgets_values": [
              "veed_clean_audio_noisy_input.wav",
              ""
          ],
          "widgets_values_named": {
              "audio": "veed_clean_audio_noisy_input.wav",
              "upload": ""
          }
      },
      {
          "id": 14,
          "type": "SaveAudioAdvanced",
          "pos": [
              2486,
              1010
          ],
          "size": [
              530,
              149.91320173284066
          ],
          "flags": [],
          "order": 3,
          "mode": 0,
          "inputs": [
              {
                  "name": "audio",
                  "type": "AUDIO",
                  "link": 8
              }
          ],
          "outputs": [
              {
                  "name": "audio",
                  "type": "AUDIO",
                  "links": null
              }
          ],
          "properties": [],
          "widgets_values": [
              "audio/VEED_Clean_Audio",
              "mp3",
              "V0"
          ],
          "widgets_values_named": {
              "filename_prefix": "audio/VEED_Clean_Audio",
              "format": "mp3",
              "quality": "V0"
          }
      },
      {
          "id": 15,
          "type": "MarkdownNote",
          "pos": [
              1050,
              1010
          ],
          "size": [
              520,
              560
          ],
          "flags": [],
          "order": 1,
          "mode": 0,
          "inputs": [],
          "outputs": [],
          "properties": [],
          "widgets_values": [
              "## VEED Clean Audio: Speech Cleanup — Workflow Guide\n\n## Input Asset\n- [veed_clean_audio_noisy_input.wav](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/veed_clean_audio_noisy_input.wav)\n\n## How to Use\n1. Load the speech recording in Load Audio.\n2. Keep strength at 0.874, or lower it to keep more room tone under the voice.\n3. Pick normalize (target -19 LUFS) or keep input level.\n\n## Limits\n- Up to 1799 seconds per run.\n- Returns 48 kHz mono; music and effects are removed with the noise.\n- Pauses are cleaned fully; peaks cap at -1.1 dBTP, so loud targets settle near -15 LUFS.\n"
          ],
          "widgets_values_named": {
              "text": "## VEED Clean Audio: Speech Cleanup — Workflow Guide\n\n## Input Asset\n- [veed_clean_audio_noisy_input.wav](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/input/veed_clean_audio_noisy_input.wav)\n\n## How to Use\n1. Load the speech recording in Load Audio.\n2. Keep strength at 0.874, or lower it to keep more room tone under the voice.\n3. Pick normalize (target -19 LUFS) or keep input level.\n\n## Limits\n- Up to 1799 seconds per run.\n- Returns 48 kHz mono; music and effects are removed with the noise.\n- Pauses are cleaned fully; peaks cap at -1.1 dBTP, so loud targets settle near -15 LUFS.\n"
          },
          "color": "#222",
          "bgcolor": "#000",
          "title": "Note: Workflow Guide"
      }
  ],
  "links": [
      [
          7,
          13,
          0,
          12,
          0,
          "AUDIO"
      ],
      [
          8,
          12,
          0,
          14,
          0,
          "AUDIO"
      ]
  ],
  "groups": [],
  "config": [],
  "extra": {
      "ds": {
          "scale": 0.9377531531912866,
          "offset": [
              -1132.151137766445,
              -586.1568342765814
          ]
      },
      "frontendVersion": "1.58.0"
  },
  "version": 0.4
188}

VEED Clean Audio：语音降噪

节点构成

官方标注的模型

工作流 JSON 复制

工作流 JSON