Name: Meta VGGT 1B 3D Scene Reconstruction API
Brand: magicapi
SKU: api-magicapi-3d-scene-attribute-inference-vufinder-vggt-1b
Availability: InStock
Author: API Market

Shashank Agarwal

# 🧊 3D Scene Reconstruction API (Meta VGGT 1B)

## Overview
VGGT (Visual Geometry Grounded Transformer) is a 1-billion-parameter 3D vision model from Meta. Its developer describes it as a feed-forward neural network that directly infers all key 3D attributes of a scene from a set of views. This listing is packaged by vufinder.

Send images or a video in `inputs`. Video is sampled every `sampling_rate` frames, and inputs are padded to one aspect ratio and resized to at most 518 pixels. The API returns one JSON file of predictions per image or sampled frame and, by default, a GLB point cloud of the scene.

## Features
- **Images or video**: `inputs` accepts JPG, JPEG, PNG and WEBP images and MP4, AVI and MOV videos.
- **Frame sampling**: `sampling_rate` (default 24) keeps every n-th video frame; the first and last frames are always included.
- **Point cloud**: `return_pcd` (default `true`) returns a GLB point cloud, built from the point head or the depth head (`pcd_source`).
- **Camera poses**: `enable_pose_postprocessing` refines the predicted camera poses by fitting unprojected depth to world points; `weighted_pose_transform` applies a weighted version of that transform.
- **Per-frame data**: `data` holds one JSON file per image or sampled frame. `to_base64` (default `true`) stores arrays as base64 strings with shape and dtype; `keys_to_exclude` drops keys you do not need.
- **Extra point attributes**: `normals` computes normals for the point cloud and `point_scales` computes a depth-based scale per point.

## Use Cases
1. **Scene capture**: turn a short video of a room, object or site into a 3D point cloud.
2. **Camera tracking**: recover camera poses for a set of photos or video frames.
3. **Robotics and mapping**: get 3D structure from camera footage without a depth sensor.
4. **3D pipelines**: use the point cloud and per-frame predictions as a starting point for meshing or further processing.

## Visual Examples
### Example Input
One short video in `inputs`, sampled every 24 frames, with normals, point scales and pose post-processing (weighted) turned on:

[Watch the example input video](https://blog.api.market/wp-content/uploads/2026/10/vggt-1b-input-api-market.mp4)

```json
{
  "inputs": [
    "https://blog.api.market/wp-content/uploads/2026/10/vggt-1b-input-api-market.mp4"
  ],
  "normals": true,
  "to_base64": true,
  "pcd_source": "point_head",
  "return_pcd": true,
  "point_scales": true,
  "sampling_rate": 24,
  "keys_to_exclude": "",
  "alpha_blend_onto": "keep",
  "weighted_pose_transform": true,
  "enable_pose_postprocessing": true
}
```

### Example Output
The request returned:

- `point_cloud`: a GLB point cloud of the scene (about 21.9 MB).
- `data`: 9 JSON files, one per sampled video frame, with the prediction arrays as base64 strings.

The GLB file is too large to preview on this page; open it in any 3D viewer that reads GLB.

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `inputs` | array of URLs | Yes | none |  | Input file URLs: JPG, JPEG, PNG or WEBP images, or MP4, AVI or MOV videos. Images and sampled video frames are padded to one aspect ratio and resized to a maximum dimension of 518 pixels. |
| `sampling_rate` | integer | No | `24` |  | For video inputs, keep every n-th frame. The first and last frames are always included. |
| `return_pcd` | boolean | No | `true` |  | Whether to return a point cloud file |
| `pcd_source` | string | No | `point_head` | `point_head`, `depth_head` | Whether the point cloud is built from the point head or the depth head output. No effect when `return_pcd` is `false`. |
| `alpha_blend_onto` | string | No | `white` | `keep`, `white`, `black`, `mean` | How to handle images with an alpha channel. `mean` blends the image onto the ImageNet mean RGB values; `keep` keeps the original pixel values. |
| `normals` | boolean | No | `false` |  | Whether to compute normals for the point cloud |
| `point_scales` | boolean | No | `false` |  | Whether to compute a scaling factor for each point based on its depth |
| `enable_pose_postprocessing` | boolean | No | `false` |  | Whether to post-process the predicted camera poses. When `true`, the poses are transformed by fitting unprojected depth to world points. |
| `weighted_pose_transform` | boolean | No | `false` |  | Whether to apply a weighted transformation to the predicted camera poses. Used when `enable_pose_postprocessing` is `true`. |
| `keys_to_exclude` | string | No | `""` |  | Comma-separated list of keys to leave out of the output JSON files |
| `to_base64` | boolean | No | `true` |  | Whether to store the arrays in the JSON files as base64 strings with shape and dtype |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.077 per API unit.
- **1 API unit = 1 successful request with up to 71 seconds of processing.** A request that takes longer is billed one more unit for each started block of 71 seconds (for example, 143 seconds of processing is 3 units, $0.231).
- More images, longer videos or a lower `sampling_rate` (more frames) take longer to process; a request uses more than one unit only when it runs past 71 seconds.
- **Example cost**: the example request above took 7.1 seconds of processing, so it used 1 API unit and cost $0.077.
- **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/3d-scene-attribute-inference-vufinder-vggt-1b`

### 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/3d-scene-attribute-inference-vufinder-vggt-1b/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "8f588e57226dc37aecdfceda935eac3ab3f8632b48d385a6c2d86cf6bf73cd23",
    "input": {
      "inputs": [
        "https://blog.api.market/wp-content/uploads/2026/10/vggt-1b-input-api-market.mp4"
      ],
      "normals": true,
      "to_base64": true,
      "pcd_source": "point_head",
      "return_pcd": true,
      "point_scales": true,
      "sampling_rate": 24,
      "keys_to_exclude": "",
      "alpha_blend_onto": "keep",
      "weighted_pose_transform": true,
      "enable_pose_postprocessing": true
    }
  }'
```

Response (201 Created):
```json
{
  "id": "nc0x2zfxq5rmy0d1d1c9vgyvv4",
  "version": "8f588e57226dc37aecdfceda935eac3ab3f8632b48d385a6c2d86cf6bf73cd23",
  "status": "starting",
  "created_at": "2026-10-23T10:40:25.176Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/3d-scene-attribute-inference-vufinder-vggt-1b/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK, `data` shortened to 3 of its 9 files):
```json
{
  "id": "nc0x2zfxq5rmy0d1d1c9vgyvv4",
  "status": "succeeded",
  "output": {
    "data": [
      "URL of the JSON file for frame 1",
      "URL of the JSON file for frame 2",
      "URL of the JSON file for frame 3"
    ],
    "point_cloud": "URL of the GLB point cloud file"
  },
  "error": null,
  "created_at": "2026-10-23T10:40:25.176Z",
  "started_at": "2026-10-23T10:41:23.724Z",
  "completed_at": "2026-10-23T10:41:30.780Z"
}
```

`output` is an object. `data` is a list of JSON file URLs, one per image or sampled frame. `point_cloud` is a GLB file URL, or `null` when `return_pcd` is `false`.

### 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 `inputs` list or a `pcd_source` other than `point_head` or `depth_head`.
- **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/3d-scene-attribute-inference-vufinder-vggt-1b";

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: "8f588e57226dc37aecdfceda935eac3ab3f8632b48d385a6c2d86cf6bf73cd23", 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({
  "inputs": [
    "https://blog.api.market/wp-content/uploads/2026/10/vggt-1b-input-api-market.mp4"
  ],
  "normals": true,
  "to_base64": true,
  "pcd_source": "point_head",
  "return_pcd": true,
  "point_scales": true,
  "sampling_rate": 24,
  "keys_to_exclude": "",
  "alpha_blend_onto": "keep",
  "weighted_pose_transform": true,
  "enable_pose_postprocessing": true
}).then((output) => console.log(output));
```

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

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/3d-scene-attribute-inference-vufinder-vggt-1b"


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": "8f588e57226dc37aecdfceda935eac3ab3f8632b48d385a6c2d86cf6bf73cd23", "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")


print(create_and_wait_for_prediction({
    "inputs": [
        "https://blog.api.market/wp-content/uploads/2026/10/vggt-1b-input-api-market.mp4"
    ],
    "normals": True,
    "to_base64": True,
    "pcd_source": "point_head",
    "return_pcd": True,
    "point_scales": True,
    "sampling_rate": 24,
    "keys_to_exclude": "",
    "alpha_blend_onto": "keep",
    "weighted_pose_transform": True,
    "enable_pose_postprocessing": True
}))
```

## FAQ
- **What can I send in `inputs`?**
  - Image URLs (JPG, JPEG, PNG, WEBP) or video URLs (MP4, AVI, MOV). Each input is padded to one aspect ratio and resized to at most 518 pixels.
- **How many frames are used from a video?**
  - Every n-th frame, set by `sampling_rate` (default 24), plus the first and last frames. The example video gave 9 frames and 9 JSON files in `data`.
- **Does it return a mesh?**
  - No. The 3D output is a GLB point cloud in `point_cloud` (about 21.9 MB in the example). Per-frame predictions are in the JSON files in `data`.
- **Should I build the point cloud from the point head or the depth head?**
  - `pcd_source` picks which model output the point cloud is built from. `point_head` is the default; try `depth_head` and compare on your own scenes.
- **How is a request billed?**
  - $0.077 per request for up to 71 seconds of processing, plus one unit for each further started block of 71 seconds. The example video took 7.1 seconds and cost $0.077.
- **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`.
- **What if my prediction fails?**
  - Failed predictions are not billed. Read the `error` field, adjust the input and try again.

## 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.


Reconstruct 3D scenes from images or video with Meta VGGT 1B: per-frame predictions as JSON and a GLB point cloud. Pay as you go, no monthly fee.