pruna

Pruna AI — Image & Video Generation

Generate images, edit photos, upscale, and create videos directly from Claude Code using Pruna AI's ultra-fast inference API.

Setup

Ensure PRUNA_API_KEY is set:

# macOS Keychain
export PRUNA_API_KEY=$(security find-generic-password -a $USER -s PRUNA_API_KEY -w)
# Or set directly
export PRUNA_API_KEY="your-key"

The MCP server must be configured in .claude/settings.local.json:

{
  "mcpServers": {
    "pruna": {
      "command": "uvx",
      "args": ["pruna-mcp-server"],
      "env": { "PRUNA_API_KEY": "${PRUNA_API_KEY}" }
    }
  }
}

Image Generation

Use generate_image with these guidelines:

Prompt structure: Subject + Style + Lighting + Composition + Quality modifiers

Aspect ratios:

• 1:1 — social media, product photos
• 16:9 — blog headers, YouTube thumbnails
• 9:16 — stories, TikTok, mobile
• 4:3 — standard photos

Model selection:

• p-image — default, fastest (~1.5s, $0.005)
• flux-dev — highest quality (~3s, $0.005)
• flux-2-klein-4b — cheapest drafts (~1s, $0.0001)

Always use English prompts. Use seed for reproducible results when iterating.

Image Editing

Use edit_image with 1-5 reference images + text instruction.

Be explicit about what to change AND what to preserve. Default aspect_ratio: "match_input_image".

Video Generation

Use generate_video with prompt structure: [Subject] + [Action] + [Camera] + [Style]

• p-video — premium quality, supports image+audio input ($0.02-0.04/s)
• wan-t2v — text-to-video only
• wan-i2v — animate a static image

Keep duration short (5s) for iterations. Use 720p for drafts, 1080p for final.

Upscaling

Use upscale_image with target 1-8 megapixels. Set enhance_realism=true for AI-generated images.