/ Verzeichnis / Playground / banana-claude
← Alle Server ↗ GitHub
● Community AgriciDaniel ⚡ Sofort

banana-claude

von AgriciDaniel · AgriciDaniel/banana-claude

AI image generation skill for Claude Code — a creative director powered by Gemini that prompts the model so you don't have to.

Banana Claude turns Claude Code into a creative director for AI images. You describe the shot at a high level; the skill expands it into a Gemini-ready prompt with composition, lighting, and style directives, then calls Gemini to render. Good for hero images, social cards, and brand-consistent stills.

▶ Demo ansehen 📦 Installieren 💡 Anwendungsfälle

Warum nutzen

Hauptfunktionen

Live-Demo

In der Praxis

banana-claude-skill.replay ▶ bereit
0/0

Installation

Repository des Autors · Offizielle Dokumentation zu Claude Skills

Anwendungsfälle

Praxisnahe Nutzung: banana-claude

How to generate a hero image for a blog post

👤 Writers and marketers who want one-shot polished hero images ⏱ ~5 min beginner

Wann einsetzen: You have a post title and a vibe in mind — you need one hero image in the next 2 minutes.

Ablauf
  1. Describe the shot
    Banana Claude: hero image for 'The 2AM Bug That Cost Us $40k' — moody, cinematic, developer at a laptop lit by monitor glow.✓ Kopiert
    → Expanded prompt with lens/lighting details echoed back
  2. Review and generate
    Generate 2 variants and save to assets/hero-2am-bug.png✓ Kopiert
    → Two PNGs with different takes on the brief

Ergebnis: A hero image that looks commissioned, not generated.

Fallstricke
  • All outputs share a 'Gemini look' — Add a style anchor — 'in the style of Annie Leibovitz portraits' — to break the default
Kombinieren mit: smart-illustrator-skill

Generate a brand-consistent social image pack

👤 Solo marketers maintaining a consistent feed look ⏱ ~15 min beginner

Wann einsetzen: You need 6-10 images that feel like they belong together.

Voraussetzungen
  • Brand style notes (palette, mood) — Write 3-5 lines; skill uses as style guardrail
Ablauf
  1. Lock a style anchor
    My brand style: muted pastels, isometric, soft shadows. Lock that as the anchor for the session.✓ Kopiert
    → Skill confirms and echoes the anchor
  2. Iterate per post
    Now generate images for these 6 post titles using the anchor.✓ Kopiert
    → 6 on-brand images

Ergebnis: A cohesive image set ready for a week of posts.

Fallstricke
  • Drift after image 3 — Re-state the anchor every 2-3 generations; Gemini's context on image specifics is fuzzy
Kombinieren mit: ai-marketing-claude-code-skill

Kombinationen

Mit anderen MCPs für 10-fache Wirkung

banana-claude-skill + smart-illustrator-skill

Banana for hero/brand images, Smart Illustrator for in-article diagrams

For posts/launch.md: banana-claude for the hero, smart-illustrator for the architecture diagrams inside.✓ Kopiert
banana-claude-skill + ai-marketing-claude-code-skill

Marketing skill plans the campaign; Banana renders the creative

Draft a 7-post launch campaign with the marketing skill, then generate hero images for each with banana-claude.✓ Kopiert

Werkzeuge

Was dieses MCP bereitstellt

WerkzeugEingabenWann aufrufenKosten
expand_brief brief: str, style_anchor?: str Before generating, to build a rich prompt ~1k Claude tokens
generate_image prompt: str, variants?: int, out_path?: str After brief expansion approved 1 Gemini call per variant
save_template name, prompt When a prompt pattern produces consistent wins 0

Kosten & Limits

Was der Betrieb kostet

API-Kontingent
Gemini free tier: 15 req/min, 1500/day
Tokens pro Aufruf
1-3k Claude tokens for prompt expansion
Kosten in €
Free on Gemini free tier; ~$0.04/image on paid tier
Tipp
Save winning prompts as templates to avoid re-paying for expansion calls.

Sicherheit

Rechte, Secrets, Reichweite

Credential-Speicherung: GEMINI_API_KEY in env
Datenabfluss: All briefs and outputs go to Google Gemini

Fehlerbehebung

Häufige Fehler und Lösungen

Gemini returns blurry/low-quality image

Include a resolution and aspect-ratio hint in the brief; ask for 'photographic, high detail, 4K'

Safety filter blocks the generation

Simplify the brief; remove ambiguous references to people or brands

Alternativen

banana-claude vs. andere

AlternativeWann stattdessenKompromiss
smart-illustrator-skillYou need diagrams and covers with position detection, not just stillsLess hand-holding on the creative brief
Midjourney / SoraYou need higher-end fidelitySeparate UI, no Claude integration

Mehr

Ressourcen

📖 Offizielle README auf GitHub lesen

🐙 Offene Issues ansehen

🔍 Alle 400+ MCP-Server und Skills durchsuchen