# Florence-2 Large - Core.Today AI API

> Microsoft's Florence-2 all-in-one vision model — captioning, object detection, phrase grounding, OCR, and segmentation in a single API. Pick a task, optionally add text input, done.

- **Provider**: Microsoft
- **Model ID**: lucataco/florence-2-large
- **Category**: Image Generation
- **Credits**: 2 per call
- **Speed**: Fast
- **Quality**: High

## Features
- One model, many tasks (task_input selection)
- Caption / Detailed Caption / Object Detection
- OCR and OCR with region boxes
- Phrase grounding and segmentation
- 3 credits per call

## Use Cases
- Unified vision pipelines with one dependency
- Detailed image captioning
- OCR with positional boxes
- Open-vocabulary detection
- Region-aware segmentation

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **image**: string - Input image

### Optional
- **task_input**: string (default: Caption) - Vision task to run (caption, detection, OCR, grounding, segmentation...)
  Options: Caption, Detailed Caption, More Detailed Caption, Caption to Phrase Grounding, Object Detection, Dense Region Caption, Region Proposal, OCR, OCR with Region
- **text_input**: string - Text for tasks that need it (grounding phrases, region coords)

## Examples

### OCR with Regions
Extract text with positional boxes
```json
{
  "model": "lucataco/florence-2-large",
  "input": {
    "image": "https://example.com/document.jpg",
    "task_input": "OCR with Region"
  }
}
```

### Detailed Caption
Get a rich multi-sentence caption
```json
{
  "model": "lucataco/florence-2-large",
  "input": {
    "image": "https://example.com/scene.jpg",
    "task_input": "More Detailed Caption"
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- One dependency covers caption+detect+OCR — simplifies vision pipelines
- 'OCR with Region' returns text plus box coordinates for layout-aware parsing
- Use 'Caption to Phrase Grounding' with text_input to locate phrases in the image
- For text-prompted detection with thresholds, Grounding DINO gives finer control

## Documentation
https://replicate.com/lucataco/florence-2-large