Introduction

The prediction API endpoints enable users to start and manage image processing sessions, such as editing, generating, and enhancing images using AI models. Users can monitor the progress of each session and retrieve the generated images or other results.

Image Generation Process

The image generation API lets users create one or more images from a text prompt. This process is asynchronous, and when a request is initiated, the response contains two important properties:

• id (uuid): Identifier of the prediction session, which can be used to query the session API to get updates on the session's progress.

• state (string): Represents the progress of the session. Initially set to scheduled, it transitions through the following states:

  state	description
  scheduled	The default state, indicating that the prediction is queued and ready for processing.
  running	The session is actively being processed by the model.
  succeed	The prediction has completed successfully.
  failed	An error occurred during the image generation, and no results are available.

Example Response snippet

{
  "id": "d61fdfd1-985e-43e9-8e7a-d776f7fb7819",
  "state": "scheduled"
}

In this example, the response shows that the prediction session has been created with the initial state set to scheduled.

To receive updates and retrieve generated images, make a request to /predict/session/{id}.

Prediction Session Results

The Prediction API response can includes both the original image uploaded by the user and the images generated as a result of the AI processing. The type property in each image object indicates whether the image is an "input" or a "result":

Example Response snippet

{
  "result": {
    "image": [
      {
        "id": 1000,
        "type": "input",
        "url": "https://app.seelab.ai/assets/original/predict/images/example-image-id"
      },
      {
        "id": 1001,
        "type": "result",
        "url": "https://app.seelab.ai/assets/original/predict/images/example-image-id"
      }
    ]
  }
}

In this example, the first image is the user's input, while the second image is a generated result.

Rate Limits

The prediction API is subject to rate limits based on the user's subscription plan. The following table shows the number of prediction requests allowed within a 10-second window for each plan: