openrouter-image-gen

name	OpenRouter Image Gen
description	Generate images using OpenRouter's API (Flux, Gemini, GPT-5 Image models). Use when Burke asks to generate, create, or make an image, thumbnail, illustration, or visual.

OpenRouter Image Generation

Generate images via OpenRouter's chat completions API. The API key is stored in ~/.max/.env as OPENROUTER_API_KEY.

Quick Start

source ~/.max/.env

curl -s https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemini-2.5-flash-image",
    "messages": [{"role": "user", "content": "YOUR PROMPT HERE"}],
    "modalities": ["image", "text"],
    "image_config": {
      "aspect_ratio": "16:9",
      "image_size": "1K"
    }
  }'

Models (ranked by recommendation)

Model	Cost	Best For
google/gemini-2.5-flash-image	Cheapest	General purpose, fast, good quality
google/gemini-3.1-flash-image-preview	Low	Extended aspect ratios, 0.5K size option
google/gemini-3-pro-image-preview	Mid	Higher quality Gemini
black-forest-labs/flux.2-pro	Mid	Photorealistic, artistic
black-forest-labs/flux.2-max	Mid	Highest quality Flux
openai/gpt-5-image-mini	Mid-high	GPT-level prompt following
openai/gpt-5-image	Expensive	Best prompt following + reasoning
bytedance-seed/seedream-4.5	Mid	ByteDance image gen

Default model: google/gemini-2.5-flash-image (cheap + good). Use Flux for photorealism, GPT-5 Image for complex prompt following.

Response Handling

Images come back as base64 data URLs in the response JSON:

{
  "choices": [{
    "message": {
      "images": [{
        "type": "image_url",
        "image_url": {
          "url": "data:image/png;base64,iVBORw0KGgo..."
        }
      }]
    }
  }]
}

Extract and save the image:

# Parse the base64 data URL from JSON response and save as PNG
echo "$RESPONSE" | jq -r '.choices[0].message.images[0].image_url.url' | sed 's|data:image/png;base64,||' | base64 -d > output.png

Send to Burke via Telegram:

curl -s -X POST http://127.0.0.1:7777/send-photo \
  -H 'Content-Type: application/json' \
  -d '{"photo": "/path/to/output.png", "caption": "Your image caption"}'

Key Parameters

• modalities: ["image", "text"] for multimodal models (Gemini, GPT-5), ["image"] for image-only (Flux, Seedream)
• aspect_ratio: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9
• image_size: 0.5K (Gemini 3.1 only), 1K (default), 2K, 4K

Image Editing (input an existing image)

Pass a source image as base64 in the user message content array:

{
  "role": "user",
  "content": [
    {"type": "text", "text": "Make the sky purple"},
    {"type": "image_url", "image_url": {"url": "data:image/png;base64,..."}}
  ]
}

Tips

• For YouTube thumbnails, use 16:9 aspect ratio and 2K or 4K size
• Write detailed, specific prompts — describe composition, colors, style, mood
• For multiple variations, make multiple API calls with slightly different prompts
• Flux models are image-only output — use "modalities": ["image"]
• Always save the PNG locally first, then send to Telegram