Name: SAM-Audio Large API: Text-Prompted Sound Isolation
Brand: magicapi
SKU: api-magicapi-sound-isolation-ai-geopti-sam-audio-large
Availability: InStock
Author: API Market

Shashank Agarwal

# 🎧 Text-Prompted Sound Isolation API (Meta SAM-Audio Large)

## Overview
SAM-Audio Large is a sound isolation model from Meta. Its developer describes it as a foundation model for isolating any sound in audio using text.

Send an audio or video file as `audio` and describe the sound you want in `description` (default `speech`). The model returns the isolated sound as an audio file. Set `output_residual` to `true` to also get the residual audio, which is everything except the target sound.

## Features
- **Text prompts**: describe the target with a simple noun phrase such as `speech`, `man speaking`, `dog barking`, `piano`, `guitar playing` or `birds chirping`.
- **Audio or video input**: `audio` accepts audio or video files (WAV, MP3, MP4 and similar). The output is audio.
- **Residual track**: `output_residual` returns a second file with everything except the target sound.
- **Span prompting**: set `use_span_prompting` to `true` and pass time ranges in `span_anchors` to mark where the sound is present (`+`) or absent (`-`).
- **Automatic span detection**: `predict_spans` detects the time spans where the target sound occurs. The schema notes it improves quality for non-ambient sounds but is slower.

## Use Cases
1. **Speech cleanup**: keep the voice from a recording or a video's soundtrack and drop the background.
2. **Instrument extraction**: pull a single instrument, such as a piano or guitar, out of a mix.
3. **Sound removal**: set `output_residual` to `true` and use the residual file to remove a sound, such as a dog barking, from a recording.
4. **Sound effect libraries**: cut clean clips of a specific sound (birds chirping, applause) from longer field recordings.

## Visual Examples
### Example Input
Audio: [Listen to the example input](https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-input-api-market.m4a)

Description: `speech`. All other fields were left at their defaults (no span prompting, no span prediction, no residual).

### Example Output
The isolated speech: [Listen to the example output](https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-output-46h86rq6-api-market.mp3)

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `audio` | string (URL) | Yes | none |  | Input audio or video file (WAV, MP3, MP4, etc.) |
| `description` | string | No | `speech` |  | Text description of the sound to isolate. Use simple noun phrases like `speech`, `man speaking`, `dog barking`, `piano`, `guitar playing`, `birds chirping` |
| `use_span_prompting` | boolean | No | `false` |  | Enable span prompting to specify time ranges where the target sound occurs. More precise but requires knowing timestamps. |
| `span_anchors` | string | No | `[]` |  | Only used when `use_span_prompting` is `true`. Time ranges as a JSON array string: `[['+', start_sec, end_sec], ...]`. `+` means the sound is present, `-` means absent. Example: `[['+', 2.0, 4.0]]` or `[['+', 1.0, 3.0], ['-', 5.0, 6.0]]` |
| `predict_spans` | boolean | No | `false` |  | Auto-detect time spans where the target sound occurs. Improves quality for non-ambient sounds but slower. |
| `output_residual` | boolean | No | `false` |  | Also output the residual audio (everything except the target sound) |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.079 per API unit.
- **1 API unit = 1 request covering up to 72 seconds of processing.** A request that takes longer uses one more unit for each started 72-second block (for example, 73 seconds of processing is 2 units, $0.158).
- Processing time, not the length of the audio, sets the number of units. `predict_spans` makes a request slower.
- **Example cost**: the example request took 4.54 seconds of processing, so it used 1 API unit = **$0.079**.
- **Failed or cancelled predictions are not billed.**
- **GET requests to check prediction status** are free and don't count toward API unit usage.
- **No monthly subscription fee.**

## API Documentation

### Base URL
All API requests go to: `https://prod.api.market/api/v1/magicapi/sound-isolation-ai-geopti-sam-audio-large`

### Authentication
Send your api.market API key in the `x-api-market-key` header:
```
x-api-market-key: YOUR_API_MARKET_KEY
```

### Asynchronous Processing
**IMPORTANT**: This API is asynchronous. The flow is:
1. Send a POST request to `/predictions`. A successful request returns **201 Created** with a prediction `id` and a `status` of `starting` or `processing`.
2. Poll GET `/predictions/{id}` with that ID.
3. Keep polling while `status` is `starting` or `processing`.
4. Stop when `status` is `succeeded` (read `output`), `failed` or `canceled` (read `error`).

A polling interval of 1 second and an overall timeout of 600 seconds (10 minutes) work well.

### Endpoints

#### Create a Prediction
```bash
curl -X POST "https://prod.api.market/api/v1/magicapi/sound-isolation-ai-geopti-sam-audio-large/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "d8a8a4fcdcbf0bdc863f6d98cd2117ec0bc02224b576c7b98b2a009a8a1f83fa",
    "input": {
      "audio": "https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-input-api-market.m4a",
      "description": "speech",
      "span_anchors": "[]",
      "predict_spans": false,
      "output_residual": false,
      "use_span_prompting": false
    }
  }'
```

Response (201 Created):
```json
{
  "id": "46h86rq68nrmt0d1d0ya87nccc",
  "version": "d8a8a4fcdcbf0bdc863f6d98cd2117ec0bc02224b576c7b98b2a009a8a1f83fa",
  "status": "starting",
  "created_at": "2026-10-23T10:09:44.168Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/sound-isolation-ai-geopti-sam-audio-large/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "46h86rq68nrmt0d1d0ya87nccc",
  "status": "succeeded",
  "output": [
    "https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-output-46h86rq6-api-market.mp3"
  ],
  "error": null,
  "created_at": "2026-10-23T10:09:44.168Z",
  "started_at": "2026-10-23T10:10:46.629Z",
  "completed_at": "2026-10-23T10:10:51.168Z"
}
```

`output` is an array of audio file URLs. With the default settings it holds one URL, the isolated target sound. With `output_residual` set to `true` it also contains the residual audio.

### Status Codes
- **201 Created**: prediction created (POST `/predictions`). This is the normal success response.
- **200 OK**: prediction details returned (GET `/predictions/{id}`).
- **400 Bad Request**: invalid input, for example a missing `audio` URL.
- **401 Unauthorized**: missing or invalid API key.
- **404 Not Found**: no prediction with that ID.
- **429 Too Many Requests**: rate limit exceeded.
- **500 Internal Server Error** or **503 Service Unavailable**: temporary server problem; retry later.

## Rate Limits
All plans allow 10 requests per second. Requests above that return `429 Too Many Requests`; back off before retrying and honour any `Retry-After` header. Polling a prediction with GET is free but still counts toward the per-second limit, so poll every 1 to 5 seconds rather than in a tight loop.

## Error Handling
- Check the HTTP status code of every request and read the response body for the error message.
- A prediction can also end with `status` set to `failed`; the `error` field then explains why. Failed predictions are not billed.
- Retry 429, 500 and 503 responses with exponential backoff. Fix the input before retrying a 400.

## Implementation Guide
### Example Code

#### JavaScript
```javascript
const apiKey = "YOUR_API_MARKET_KEY";
const baseUrl = "https://prod.api.market/api/v1/magicapi/sound-isolation-ai-geopti-sam-audio-large";

async function createAndWaitForPrediction(input) {
  const response = await fetch(`${baseUrl}/predictions`, {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "x-api-market-key": apiKey
    },
    body: JSON.stringify({ version: "d8a8a4fcdcbf0bdc863f6d98cd2117ec0bc02224b576c7b98b2a009a8a1f83fa", input })
  });

  if (response.status !== 201) {
    throw new Error(`API error: ${response.status} ${await response.text()}`);
  }

  const prediction = await response.json();
  return pollForResults(prediction.id);
}

const delay = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

async function pollForResults(predictionId) {
  const maxTimeout = 600000;
  const pollInterval = 1000;
  const startTime = Date.now();

  while (Date.now() - startTime < maxTimeout) {
    const response = await fetch(`${baseUrl}/predictions/${predictionId}`, {
      headers: { "x-api-market-key": apiKey }
    });

    if (!response.ok) {
      throw new Error(`API error: ${response.status} ${await response.text()}`);
    }

    const prediction = await response.json();

    if (prediction.status === "succeeded") {
      return prediction.output;
    }

    if (prediction.status === "failed" || prediction.status === "canceled") {
      throw new Error(`Prediction ${prediction.status}: ${prediction.error || "Unknown error"}`);
    }

    await delay(pollInterval);
  }

  throw new Error("Prediction timed out after 10 minutes");
}

createAndWaitForPrediction({
    "audio": "https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-input-api-market.m4a",
    "description": "speech",
    "span_anchors": "[]",
    "predict_spans": false,
    "output_residual": false,
    "use_span_prompting": false
  }).then((output) => console.log(JSON.stringify(output, null, 2)));
```

#### Python
```python
import json
import time
import requests

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/sound-isolation-ai-geopti-sam-audio-large"


def create_and_wait_for_prediction(input_data):
    headers = {"Content-Type": "application/json", "x-api-market-key": API_KEY}
    response = requests.post(
        f"{BASE_URL}/predictions",
        headers=headers,
        json={"version": "d8a8a4fcdcbf0bdc863f6d98cd2117ec0bc02224b576c7b98b2a009a8a1f83fa", "input": input_data},
    )

    if response.status_code != 201:
        raise Exception(f"API error: {response.status_code} {response.text}")

    return poll_for_results(response.json()["id"])


def poll_for_results(prediction_id):
    max_timeout = 600
    poll_interval = 1
    headers = {"x-api-market-key": API_KEY}
    start_time = time.time()

    while time.time() - start_time < max_timeout:
        response = requests.get(f"{BASE_URL}/predictions/{prediction_id}", headers=headers)

        if response.status_code != 200:
            raise Exception(f"API error: {response.status_code} {response.text}")

        prediction = response.json()

        if prediction["status"] == "succeeded":
            return prediction["output"]

        if prediction["status"] in ("failed", "canceled"):
            raise Exception(f"Prediction {prediction['status']}: {prediction.get('error') or 'Unknown error'}")

        time.sleep(poll_interval)

    raise Exception("Prediction timed out after 10 minutes")


output = create_and_wait_for_prediction({
    "audio": "https://blog.api.market/wp-content/uploads/2026/10/sam-audio-large-input-api-market.m4a",
    "description": "speech",
    "span_anchors": "[]",
    "predict_spans": False,
    "output_residual": False,
    "use_span_prompting": False
})
print(json.dumps(output, indent=2))
```

## FAQ
- **How should I describe the sound?**
  - Use a short noun phrase such as `speech`, `man speaking`, `dog barking` or `piano`. The default is `speech`.
- **Can I get the audio without the target sound?**
  - Yes. Set `output_residual` to `true`; `output` then also contains the residual audio, which is everything except the target sound.
- **When should I use span prompting?**
  - When you know where the sound occurs. Set `use_span_prompting` to `true` and pass `span_anchors`, for example `[['+', 2.0, 4.0]]` for a sound present from 2 to 4 seconds. If you do not know the timestamps, try `predict_spans` instead.
- **Can I send a video file?**
  - Yes. `audio` accepts video files such as MP4. The output is an audio file.
- **How is a request billed?**
  - $0.079 per request for up to 72 seconds of processing, plus one more unit for each further started 72-second block. Failed requests are not billed.
- **Is the 201 response from POST an error?**
  - No. 201 Created means the prediction was accepted. Use the returned `id` to poll GET `/predictions/{id}` until `status` is `succeeded`, `failed` or `canceled`.

## Contact Us
If you have any questions, reach out to us through the chat on the bottom right of the page. You can also email our support team at hello@magicapi.com.


Isolate any sound in an audio or video file with Meta's SAM-Audio Large by describing it in text, with an optional residual track.