Skip to main content
VEED Clean Audio isolates the voice in a recording. It strips background noise, and because it is built for speech, music and sound effects go with the noise: what comes back is the dialogue on its own, cleaned and ready to use. In ComfyUI the model runs through two partner nodes. VEED Clean Audio takes a standalone recording and returns cleaned audio. VEED Clean Audio (in Video) takes a clip, cleans only its soundtrack, and returns the video with the picture untouched. Both are cloud API nodes: they need a Comfy account with credits and download no model files locally. The two nodes return different things. Clean Audio returns 48 kHz mono audio. Clean Audio (in Video) returns the video itself, with the picture untouched and its soundtrack replaced by that cleaned 48 kHz mono speech. Both accept runs up to 1799 seconds, just under 30 minutes. For lip sync, talking portraits, and background removal, see VEED Video Nodes.
To use the Partner Nodes, you need to ensure that you are logged in properly and using a permitted network environment. Please refer to the Partner Nodes Overview section of the documentation to understand the specific requirements for using the Partner Nodes.
Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Available workflows

​
Clean Audio

Clean a voice recording before it goes anywhere else. The node is the first step of a voice pipeline: send the cleaned speech to a lip sync node, a transcription step, or a mix, and the noise floor of the original never reaches them. VEED Clean Audio: Speech Cleanup workflow preview

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “VEED Clean Audio: Speech Cleanup” in Template Library
Input material

veed_clean_audio_noisy_input.wav

Load into the LoadAudio node

​
Clean Audio (in Video)

Clean the speech inside a clip without touching the picture. Only the soundtrack is replaced, with the cleaned 48 kHz mono speech, so the result is the same edit with a usable voice track. VEED Clean Audio: Video Speech Cleanup workflow preview

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “VEED Clean Audio: Video Speech Cleanup” in Template Library
Input material

veed_clean_audio_input_h3_omniref_person_dialogue_plus_noise.mp4

Load into the LoadVideo node. The clip needs an audio track
A clip with no audio track stops with an error instead of returning a silent video. Only the first audio track is kept: extra audio tracks and embedded subtitles are dropped, so export those first when the clip carries more than one.

​
Node controls

Pauses between words are always cleaned completely, whatever strength is set to, and values above the default change little. Peaks are capped at -1.1 dBTP, so a loud target stops early: typical speech settles near -15 LUFS.

​
Tips

  • Leave strength at the default first. It is tuned for speech. Lower it only when the room tone behind the voice is part of the sound you want to keep.
  • Normalize when the cleaned audio has to match other clips. A shared target_lufs is what makes a series of recordings sit at the same level.
  • Do not expect music back. Music and effects are removed along with the noise, so this node cleans dialogue, not a full mix. Run it on the voice stem when a soundtrack has to survive.
  • Cut long recordings before the run. The cap is 1799 seconds, and a shorter file is cheaper at the per-minute rate.
  • Chain it into the talking nodes. Cleaned speech makes a better input for lip sync and talking portraits than the raw recording.

​
Get started

  1. Update ComfyUI to the latest version
  2. Go to Template Library, search for VEED