transloadit-media-processing

The full skill.

Original instructions from the publisher’s SKILL.md

# Transloadit Media Processing

Process, transform, and encode media files using Transloadit's cloud infrastructure.
Supports video, audio, images, and documents with 86+ specialized processing robots.

## When to Use This Skill

Use this skill when you need to:

- Encode video to HLS, MP4, WebM, or other formats
- Generate thumbnails or animated GIFs from video
- Resize, crop, watermark, or optimize images
- Convert between image formats (JPEG, PNG, WebP, AVIF, HEIF)
- Extract or transcode audio (MP3, AAC, FLAC, WAV)
- Concatenate video or audio clips
- Add subtitles or overlay text on video
- OCR documents (PDF, scanned images)
- Run speech-to-text or text-to-speech
- Apply AI-based content moderation or object detection
- Build multi-step media pipelines that chain operations together

## Setup

### Option A: MCP Server (recommended for Copilot)

Add the Transloadit MCP server to your IDE config. This gives the agent direct access
to Transloadit tools (`create_template`, `create_assembly`, `list_assembly_notifications`, etc.).

**VS Code / GitHub Copilot** (`.vscode/mcp.json` or user settings):

```json
{
  "servers": {
    "transloadit": {
      "command": "npx",
      "args": ["-y", "@transloadit/mcp-server", "stdio"],
      "env": {
        "TRANSLOADIT_KEY": "YOUR_AUTH_KEY",
        "TRANSLOADIT_SECRET": "YOUR_AUTH_SECRET"
      }
    }
  }
}
```

Get your API credentials at https://transloadit.com/c/-/api-credentials

### Option B: CLI

If you prefer running commands directly:

```bash
npx -y @transloadit/node assemblies create \
  --steps '{"encoded": {"robot": "/video/encode", "use": ":original", "preset": "hls-1080p"}}' \
  --wait \
  --input ./my-video.mp4
```

## Core Workflows

### Encode Video to HLS (Adaptive Streaming)

```json
{
  "steps": {
    "encoded": {
      "robot": "/video/encode",
      "use": ":original",
      "preset": "hls-1080p"
    }
  }
}
```

### Generate Thumbnails from Video

```json
{
  "steps": {
    "thumbnails": {
      "robot": "/video/thumbs",
      "use": ":original",
      "count": 8,
      "width": 320,
      "height": 240
    }
  }
}
```

### Resize and Watermark Images

```json
{
  "steps": {
    "resized": {
      "robot": "/image/resize",
      "use": ":original",
      "width": 1200,
      "height": 800,
      "resize_strategy": "fit"
    },
    "watermarked": {
      "robot": "/image/resize",
      "use": "resized",
      "watermark_url": "https://example.com/logo.png",
      "watermark_position": "bottom-right",
      "watermark_size": "15%"
    }
  }
}
```

### OCR a Document

```json
{
  "steps": {
    "recognized": {
      "robot": "/document/ocr",
      "use": ":original",
      "provider": "aws",
      "format": "text"
    }
  }
}
```

### Concatenate Audio Clips

```json
{
  "steps": {
    "imported": {
      "robot": "/http/import",
      "url": ["https://example.com/clip1.mp3", "https://example.com/clip2.mp3"]
    },
    "concatenated": {
      "robot": "/audio/concat",
      "use": "imported",
      "preset": "mp3"
    }
  }
}
```

## Multi-Step Pipelines

Steps can be chained using the `"use"` field. Each step references a previous step's output:

```json
{
  "steps": {
    "resized": {
      "robot": "/image/resize",
      "use": ":original",
      "width": 1920
    },
    "optimized": {
      "robot": "/image/optimize",
      "use": "resized"
    },
    "exported": {
      "robot": "/s3/store",
      "use": "optimized",
      "bucket": "my-bucket",
      "path": "processed/${file.name}"
    }
  }
}
```

## Key Concepts

- **Assembly**: A single processing job. Created via `create_assembly` (MCP) or `assemblies create` (CLI).
- **Template**: A reusable set of steps stored on Transloadit. Created via `create_template` (MCP) or `templates create` (CLI).
- **Robot**: A processing unit (e.g., `/video/encode`, `/image/resize`). See full list at https://transloadit.com/docs/transcoding/
- **Steps**: JSON object defining the pipeline. Each key is a step name, each value configures a robot.
- **`:original`**: Refers to the uploaded input file.

## Tips

- Use `--wait` with the CLI to block until processing completes.
- Use `preset` values (e.g., `"hls-1080p"`, `"mp3"`, `"webp"`) for common format targets instead of specifying every parameter.
- Chain `"use": "step_name"` to build multi-step pipelines without intermediate downloads.
- For batch processing, use `/http/import` to pull files from URLs, S3, GCS, Azure, FTP, or Dropbox.
- Templates can include `${variables}` for dynamic values passed at assembly creation time.