Separate Voice API

Pricing

25 credits per request

Fixed cost regardless of audio length or file size.

Overview

The Separate Voice API separates vocals (voice) and instrumental (background music/effects) from audio or video files using dedicated AI models for high-quality audio separation.

Endpoint: https://api.revidapi.com/paid/separate/audio

Endpoints

Method	Endpoint	Auth	Description
POST	`https://api.revidapi.com/paid/separate/audio`	✅ Required	Create voice separation task
GET	`https://edit.revidapi.com/api/get/{task_id}`	❌	Check task status
GET	`https://edit.revidapi.com/output/{filename}`	❌	Download result file

Request

Headers

x-api-key: Required
Content-Type: application/json (JSON body) or multipart/form-data (form body)

Parameters

Parameter	Type	Description
`audio_url`	string	Required. URL of audio/video file to separate
`output_format`	string	`mp3`, `wav`, `flac`. Default: `mp3`
`headers`	object	Custom headers when server downloads file (cookie, Referer, ...)

Note: Supports both JSON body and multipart/form-data. In n8n you can use form-data with field audio_url.

Output Formats

Format	Quality	Size	Description
`mp3`	Good	Small	Common format, high compatibility
`wav`	Highest	Large	Uncompressed, original quality
`flac`	High	Medium	Lossless compression

Response

Immediate Response (Task Created)

POST endpoint returns immediately with task_id:

{
  "task_id": "60174a16-b6fa-47b1-8c5f-c54cc434c0df",
  "status": "pending",
  "message": "Task queued",
  "type": "separate_voice"
}

Task Status (GET)

Use task_id to poll status:

GET https://edit.revidapi.com/api/get/{task_id}

⏳ Processing

{
  "task_id": "60174a16-b6fa-47b1-8c5f-c54cc434c0df",
  "status": "processing",
  "type": "separate_audio",
  "progress": 50,
  "message": "Processing with UVR-MDX-NET-Voc_FT.onnx...",
  "created_at": "2025-12-24T05:41:40.670179",
  "completed_at": null,
  "result": null
}

✅ Completed

{
  "task_id": "60174a16-b6fa-47b1-8c5f-c54cc434c0df",
  "status": "completed",
  "type": "separate_voice",
  "progress": 100,
  "message": "Voice separation completed",
  "created_at": "2025-12-24T05:41:40.670179",
  "completed_at": "2025-12-24T05:42:27.137799",
  "result": {
    "vocals_url": "/output/vocals_60174a16.mp3",
    "instrumental_url": "/output/instrumental_60174a16.mp3",
    "model_used": "UVR-MDX-NET-Voc_FT.onnx"
  }
}

❌ Failed

{
  "task_id": "60174a16-b6fa-47b1-8c5f-c54cc434c0df",
  "status": "failed",
  "type": "separate_voice",
  "progress": 0,
  "message": "Failed to download audio: HTTP 401",
  "created_at": "2025-12-24T05:41:40.670179",
  "completed_at": "2025-12-24T05:41:41.123456",
  "result": null
}

Result Fields

Field	Type	Description
`vocals_url`	string	URL of vocals file (relative path)
`instrumental_url`	string	URL of instrumental/background file (relative path)
`model_used`	string	Model that was used

Note: URLs are relative paths, need to add domain https://edit.revidapi.com to download.

Example Requests

Example

JSON body:

curl -X POST https://api.revidapi.com/paid/separate/audio \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_API_KEY" \
  -d '{"audio_url": "https://example.com/audio.mp3"}'

Multipart (n8n): Use contentType: multipart-form-data, field audio_url = audio URL (e.g. from previous response).

Check Status

curl https://edit.revidapi.com/api/get/60174a16-b6fa-47b1-8c5f-c54cc434c0df

Download Results

# Download vocals
curl -O https://edit.revidapi.com/output/vocals_60174a16.mp3

# Download instrumental
curl -O https://edit.revidapi.com/output/instrumental_60174a16.mp3

Processing Time

Audio Length	Fast Model	Quality Model
1 minute	~30s	~60s
5 minutes	~2 minutes	~4 minutes
10 minutes	~4 minutes	~8 minutes

Workflow

┌─────────────────────────────────────┐
│ 1. POST separate-voice              │
│    Input: audio_url, model          │
│    Output: task_id                  │
└─────────────────────────────────────┘
                ↓
┌─────────────────────────────────────┐
│ 2. Wait 30-60 seconds               │
│    (depending on audio length)      │
└─────────────────────────────────────┘
                ↓
┌─────────────────────────────────────┐
│ 3. GET /api/get/{task_id}           │
│    Check status                      │
└─────────────────────────────────────┘
                ↓
        ┌──────┴──────┐
        │             │
   processing     completed
        │             │
        ↓             ↓
   Loop back    ┌─────────────────────┐
   to Wait      │ 4. Download files   │
                │    - vocals_url     │
                │    - instrumental_url│
                └─────────────────────┘

Use Cases

1. Create Karaoke Videos

Original video → Separate instrumental → Combine with new vocals

2. Translate Videos While Keeping Background Music

Original video → Separate instrumental + vocals
→ Translate vocals → TTS in new language
→ Combine TTS + instrumental

3. Remove Background Music, Keep Speech

Tutorial video → Separate vocals → Use clean vocals

4. Extract Background Music

Video → Separate instrumental → Use as background music

Important Notes

1. Audio URL

URL must be public or provide headers with cookie/token
URLs from TikTok/Douyin often expire quickly (few minutes) - should download first
Supports both video and audio: MP4, MP3, WAV, FLAC, WebM, etc.

2. Separation Quality

Vocals: Main voice/singing
Instrumental: Background music, effects, sounds

Tips for better quality:

Higher quality source audio → better separation results
If vocals still mixed in, can run twice (2-pass)

3. Limits

Maximum file size: 500MB
Maximum duration: 30 minutes
Timeout: 10 minutes per task

Troubleshooting

Error: HTTP 401/403

Cause: URL requires authentication

Solution:

{
  "audio_url": "https://protected-site.com/audio.mp3",
  "headers": {
    "Cookie": "session=abc123"
  }
}

Error: File too small

Cause: Download failed, file corrupted

Solution: - Check if URL is valid - Try downloading manually to see if it works - URL may have expired

Error: Model not found

Cause: Incorrect model name

Instrumental still contains vocals

Cause: Model cannot separate perfectly 100%

Solution: 1. Use higher quality model: UVR-MDX-NET-Voc_FT.onnx 2. Run twice (2-pass): Separate once → Get instrumental → Separate again

Usage Notes

Fixed Pricing: This endpoint charges a fixed 25 credits per request, regardless of audio length.
POST returns immediately with task_id (does not wait for processing)
GET task status to retrieve results (poll in loop)
Supported formats: MP3, WAV, FLAC, MP4, WebM, and other common audio/video formats
Asynchronous processing: Use webhooks for better reliability with long audio files
Model selection: Choose model based on quality vs speed requirements

Best Practices

Use Webhooks: Always use webhooks for better reliability, especially for long audio files
Unique IDs: Provide unique id values for tracking requests
Model Selection: Use quality model for final production, fast model for testing
URL Accessibility: Ensure audio URLs are publicly accessible or provide proper authentication headers
File Size: Consider file size and duration limits before submitting requests
Error Handling: Implement robust error handling for various HTTP status codes