# CogVLM2 Video - Core.Today AI API

> Video understanding and captioning — ask free-form questions about a video and get detailed answers about actions, scenes, and content. Great for video search indexing and moderation prep.

- **Provider**: CogVLM
- **Model ID**: chenxwh/cogvlm2-video
- **Category**: Video Generation
- **Credits**: 28 per video query
- **Speed**: Medium
- **Quality**: High

## Features
- Free-form questions about video content
- Action and scene understanding
- Detailed multi-sentence answers
- Temperature/top_p generation controls

## Use Cases
- Video search index descriptions
- Content moderation pre-screening
- Automatic video summaries
- Accessibility descriptions

## 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
- **input_video**: string - Video to analyze

### Optional
- **prompt**: string (default: Describe this video.) - Question or instruction about the video
- **top_p**: number (default: 0.1) - When decoding text, samples from the top p percentage of most likely tokens; lower to ignore less likely tokens
- **temperature**: number (default: 0.1) - Sampling temperature
- **max_new_tokens**: integer (default: 2048) - Maximum answer length

## Examples

### Video Summary
Get a detailed description of what happens in a clip
```json
{
  "model": "chenxwh/cogvlm2-video",
  "input": {
    "input_video": "https://example.com/clip.mp4",
    "prompt": "Describe this video in detail, including actions and scene changes."
  }
}
```

## 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
- Ask targeted questions ('What products appear?') for extraction workflows
- Keep temperature low (0.1) for factual, consistent descriptions
- Pair with multilingual-e5-large to build searchable video indexes

## Documentation
https://replicate.com/chenxwh/cogvlm2-video