model-bridge

Model Bridge Plugin

Expert-in-the-middle prompting for AI media generation models

Claude helps you translate everyday ideas into model-specific technical prompts that leverage each AI model's unique strengths and syntax.

---

What This Plugin Does

You speak in natural language about what you want to create. Claude translates it into optimized prompts for:

• FLUX.1 - Premium image generation with excellent text rendering
• SDXL - Fast iteration with extensive LoRA ecosystem
• LTX Video - Text-to-video with cinematic camera control
• Imagen 4 - Multi-narrative editorial illustrations
• Gemini - Multimodal AI for vision and evaluation

Educational Focus: This plugin teaches you prompt patterns, not just generates them. You'll learn how different models work and develop intuition over time.

---

Installation

Prerequisites

• Claude Code CLI
• (Optional) Local models: Apple Silicon Mac, Python 3.10+, MLX framework
• (Optional) Cloud models: Google Cloud account with Vertex AI enabled

Install Plugin

claude install model-bridge

Or for local development:

git clone https://github.com/lando-labs/claude-plugins
claude --plugin-dir ./claude-plugins/model-bridge

---

Quick Start

Generate a FLUX Prompt

You: I want to create a coffee shop image with the text "DAWN BREW" on it

Claude: FLUX.1 is perfect for this - it excels at text rendering.
        Here's an optimized FLUX prompt:

        Minimalist coffee packaging design on a marble surface, the text
        "DAWN BREW" in elegant copper foil sans-serif typography at center,
        steam rising from a cup in soft focus background, warm morning light
        from the left, color palette #8B4513 brown and #D4AF37 gold accents,
        commercial product photography

        Why this works:
        - Quotation marks around text ensure FLUX renders it correctly
        - Front-loaded subject (coffee packaging)
        - HEX codes for color precision
        - Camera specs implied through "commercial product photography"

Generate an SDXL Prompt

You: Create a fantasy landscape, fast iteration

Claude: SDXL is ideal for rapid iteration (~8s generation).
        Here's an SDXL prompt pair:

        Positive:
        Vast mountain valley at dawn, ancient stone ruins covered in moss,
        mist rolling through pine forests, snow-capped peaks catching golden
        light, epic scale, cinematic composition, volumetric god rays, fantasy
        concept art style, 8K ultra detailed

        Negative:
        low quality, blurry, jpeg artifacts, text, watermark, modern buildings

        This uses SDXL's dual-prompt system to include what you want and
        exclude what you don't.

Generate a Video Prompt

You: Create a nature video of a fox

Claude: For LTX Video, we need to choreograph motion. Here's a 12-second
        shot note prompt:

        A red fox emerges from tall golden grass at dawn. The camera tracks
        alongside at a low angle as the fox moves through the field. Soft
        morning mist catches the warm backlight creating a halo effect around
        the fur. The fox pauses, ears perked, then continues walking. Handheld
        documentary style with subtle natural shake, shallow depth of field on
        85mm lens, warm color grading.

        This structures motion over time with explicit camera movement.

---

Skills Overview

Skill	Model	Use When
flux-prompt	FLUX.1	Text in images, photorealism, natural language
sdxl-prompt	SDXL	Speed, LoRAs, negative prompts, iteration
ltx-video-prompt	LTX Video	Short clips, camera movement, motion
imagen-prompt	Imagen 4	Editorial compositions, multi-story synthesis
vertex-setup	Vertex AI	Setting up Google Cloud for Imagen/Gemini

---

Model Comparison

Feature	FLUX.1	SDXL	LTX Video	Imagen 4
Speed	20-80s	~8s	~17s	Varies
Quality	Excellent	Good	Good	Excellent
Text rendering	Excellent	Weak	N/A	Good
Negative prompts	No	Yes	No	No
Ecosystem	Limited	Massive	New	Cloud-only
Infrastructure	Local/Cloud	Local/Cloud	Local	Cloud (GCP)

Choose based on:

• FLUX when quality and text rendering matter most
• SDXL when speed and iteration matter most
• LTX Video when you need motion and camera control
• Imagen when you need multi-narrative editorial compositions

---

Usage Examples

Example 1: Portrait Photography

You: Photorealistic portrait of a woman

Claude activates flux-prompt skill: