Generate Image

Generate Image is a skill in K-Dense's Claude Scientific Writer (MIT) for photos, illustrations, concept art, logos, and image edits. Version 2.0 (reviewed 2026-07-26) targets the OpenRouter Image API (POST /api/v1/images), which puts Gemini, FLUX, Seedream, Recraft, GPT-Image, and roughly thirty other models behind one request shape. Earlier listings mentioned only FLUX.2 Pro and Gemini 3 Pro.

Key Features

• One script, many models: scripts/generate_image.py defaults to google/gemini-3.1-flash-image. The skill's model table suggests google/gemini-3-pro-image for the top Gemini tier, black-forest-labs/flux.2-pro for photoreal control and seeds, flux.2-klein-4b for cheap iteration, recraft/recraft-v4-vector for SVG, and openai/gpt-image-2 for transparent backgrounds.
• Parameter awareness: models reject parameters they do not support, so the script sends only flags you pass. For example, --resolution works for Gemini but not FLUX, and --seed works for FLUX but not Gemini or OpenAI. --list-models prints live support and needs no key.
• Editing and references: -i is repeatable and accepts local files, URLs, or data URLs for edits and composites. Reference limits vary by model, from 1 (Recraft) to 16 (OpenAI).
• Cost visibility: the per-request cost from usage.cost is printed after each call. Billing is all or nothing: a failed generation is not charged.

Use Cases

• Hero images and backgrounds for posters and slides.
• Conceptual illustrations for a paper or blog post.
• Vector logos and quick variations of a visual idea.

Pricing and Access

The skill is free (MIT) and uses only the Python standard library (3.9+). Every generation is a paid OpenRouter request under your OPENROUTER_API_KEY; the price depends on the model you pick.

Getting Started

1. Install the plugin in Claude Code: /plugin marketplace add https://github.com/K-Dense-AI/claude-scientific-writer, then /plugin install claude-scientific-writer.
2. Set OPENROUTER_API_KEY in your environment or an ignored .env file.
3. Run python scripts/generate_image.py "A beautiful sunset over mountains", or edit with python scripts/generate_image.py "Make the sky purple" -i photo.jpg -o edited.png.

Limitation: reference images are uploaded to OpenRouter, so do not send unpublished or sensitive data. Sending an unsupported parameter returns an HTTP 400 rather than being ignored.

Frequently Asked Questions

Should I use it for flowcharts?

No. The skill points technical diagrams to Scientific Schematics, which adds a quality review loop.

Can I get several images at once?

Yes, with --n on models that allow it: Seedream and OpenAI up to 10, Recraft up to 6, Gemini and FLUX only 1.

Alternatives

• Nano Banana: Google's image model, usable directly without the script.
• GPT Image 2: OpenAI's image model, also reachable through this skill.
• Canvas Design Skill: Anthropic's skill for designed visual pieces rather than raw generation.

Conclusion

Generate Image is a thin, well-documented wrapper that lets Claude call almost any image model on OpenRouter from one command. Choose cheap models while iterating, and keep sensitive references local. See more in the skills hub.