My Krea

Krea AI Generation

Generate images and videos via the official Krea CLI (@krea-ai/cli).

⚡ Quick Start

# Generate an image (async by default)
krea generate image -p "a cyberpunk cityscape at night" -m bfl/flux-1-dev --width 1024 --height 1024

# Generate and download immediately
krea generate image -p "..." -o ./output.png

# Generate a video
krea generate video -p "ocean waves crashing at golden hour" -m hailuo/mini-max --duration 5

# Check a job status
krea jobs wait <job_id>

Prerequisites

• Install the CLI: npm install -g @krea-ai/cli
• Set KREA_API_KEY in your environment, or run krea auth login to store credentials.
• Verify with krea doctor

Core Workflow

1. Discover available commands

krea --help
krea generate image --help
krea models list          # list all models
krea models list --json | jq '.[] | select(.category=="image") | .id'

2. Generate an image

krea generate image -p "a cyberpunk cityscape at night" -m bfl/flux-1-dev --width 1024 --height 1024

Image models: bfl/flux-1-dev (default), bfl/flux-1.1-pro, bfl/flux-1.1-pro-ultra, google/imagen-4, google/nano-banana, ideogram/ideogram-3, openai/gpt-image-2, bytedance/seedream-5-lite, luma/uni-1

Returns: {"job_id": "...", "status": "completed", "urls": ["https://..."]}

3. Generate a video

krea generate video -p "ocean waves crashing at golden hour" -m hailuo/mini-max --duration 5 --aspect 16:9

Video models: hailuo/mini-max, kling/kling-2, runway/gen-4, pika/pika-2, google/veo-3, luma/dream-machine, openai/sora

Aspect ratios: 16:9, 9:16, 1:1

4. Poll for results / download

All generations are async unless you use --wait or -o.

# Block until complete and print URL
krea generate image -p "..." --wait

# Block and save to file (implies --wait)
krea generate image -p "..." -o ./result.png

# Or manually poll
krea jobs wait <job_id>

Completed jobs include urls with the generated media URL. When using -o, the file is saved locally and the path is returned as saved_path.

5. Check recent jobs

krea jobs list --limit 10

Before Generating — Checklist

• Model: Which model matches the user's quality/speed needs?
• Dimensions: Default is 1024×1024 (images) and 16:9 5s (video).
  • bfl/flux-1.1-pro max width: 1440px
• Source image: For image-to-image, pass --image <url>.
• Async nature: Without --wait or -o, the job is queued and a job_id is returned. Offer to check status.

Assets & Uploads

Upload an asset (local file)

krea upload ./photo.jpg --name "source-photo"

The returned URL can then be passed to --image for image-to-image or video generation.

Anti-Patterns & Gotchas

Response Format for Users

After generation (async):

Queued job 8089cf88-... (type: flux, status: scheduled). I'll check the result...

After generation (with -o or --wait):

✅ Done! Saved to ./result.png (also available at https://gen.krea.ai/images/....png)

If failed:

❌ Job failed: {status} — check krea jobs list for details.

Examples

# Image with specific model and output file
krea generate image -p "neon samurai in rain" -m bfl/flux-1.1-pro --width 1440 --height 480 -o ./banner.png

# Video from image
krea generate video -p "camera slowly zooms out" --image "https://gen.krea.ai/assets/..." -m kling/kling-2 --duration 5

# Poll loop for a known job
krea jobs wait 5d95517b-3c44-4a38-95eb-1bbdb3dbf118

Troubleshooting

Migration from mcp2cli

If you previously used mcp2cli @krea, migrate to the native CLI:

npm install -g @krea-ai/cli
rm ~/.local/bin/krea   # remove old mcp2cli wrapper if present

Key syntax changes:

• krea generate-image → krea generate image
• krea get-job → krea jobs wait
• krea list-jobs → krea jobs list
• krea upload-asset → krea upload

Related Skills

• @skills/my-web-search-kagi — For searching reference images or style inspiration online
• @skills/my-tech-stack — For tool recommendations if extending the pipeline
• @skills/my-workflow — For commit discipline when saving generated assets to version control