> Documentation index: https://api.veed.io/llms.txt. Fetch it to find every other page, as Markdown.

# Submit a clean audio (video) job

Model: Clean Audio Video
Endpoint: `POST /v1/clean-audio-video`

Start an asynchronous job that removes background noise from speech, with the same model that powers Clean Audio in the VEED editor. Keeps the picture as it is and replaces only the audio track.

It is built for speech, in any language. Music and sound effects count as noise and are removed. The video stream is copied, not re-encoded, so its frames, resolution and bitrate are the input's.

**Inputs**

- `video_url` — public URL of the video; MP4, MOV, MKV or WebM with an H.264, HEVC, AV1, MPEG-4, VP8 or VP9 video stream and an audio track, up to 30 minutes and 2 GB. Only the first audio track is kept: other audio tracks and subtitles are dropped. Split anything longer into separate jobs
- `strength` — optional; how much of the original may remain under speech, between `0` and `1`. Lower keeps more room tone behind the voice
- `target_lufs` — optional; the output's integrated loudness, between `-40` and `-8` LUFS
- `normalize_loudness` — optional; `false` keeps the input level instead of normalizing it

The result is an MP4 with AAC audio for H.264, HEVC, AV1 or MPEG-4 input, and a WebM with Opus audio for VP8 or VP9 input. The cleaned track is 48 kHz mono.

**What happens next**

The job is **accepted immediately** — you get `202 Accepted` with a `job_id` and status `PROCESSING`. Processing takes roughly 0.2–0.5× the audio's duration; poll `GET /v1/clean-audio-video/{job_id}` until the job is `COMPLETED` or `FAILED`.

## Request body

- `normalize_loudness` · boolean · optional — Set to false to skip loudness normalization and keep the input level. Defaults to `true`.
- `strength` · number · optional — How much of the original is allowed to remain under speech: the suppression floor is 1 - strength. Lower keeps more room tone behind the voice; silence between words is always fully cleaned. Between 0 and 1. Defaults to `0.874`.
- `target_lufs` · number · optional — Integrated loudness of the output in LUFS (ITU-R BS.1770); true peak is capped at -1.1 dBTP. Ignored when normalize_loudness is false. A null reads as omitted: set normalize_loudness to false to skip normalization. Between -40 and -8. Defaults to `-19`.
- `video_url` · url · required — URL of the video to clean: MP4, MOV, MKV or WebM with an H.264, HEVC, AV1, MPEG-4, VP8 or VP9 video stream and an audio track, up to 30 minutes and 2 GB. Only the first audio track is kept.

## Parameters

- `X-Veed-Store-IO` (header) — Set to `0` to not store this request's and response's bodies. They are then not available in your request logs for debugging.
- `X-Veed-Media-Expiration-Seconds` (header) — Number of seconds before the media URLs returned for this request expire. A value above the maximum is capped rather than rejected.

## Example request

```bash
curl -X POST "https://api.veed.io/v1/clean-audio-video" \
  -H "Authorization: Bearer $VEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "video_url": "https://static-assets.veed.io/api-examples/clean-audio-video-input.mp4"
}'
```

## Example response (202)

```json
{
  "data": {
    "job_id": "123e4567-e89b-12d3-a456-426614174000",
    "status": "PROCESSING",
    "credits_estimated": 2
  }
}
```

Rendered page: https://api.veed.io/docs/api/post-v1-clean-audio-video