From model-bridge
Generate optimized prompts for SDXL and execute fast image generation via MLX.
How this skill is triggered — by the user, by Claude, or both
Slash command
/model-bridge:sdxl-promptThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Generate optimized prompts for SDXL and execute fast image generation via MLX.
Generate optimized prompts for SDXL and execute fast image generation via MLX.
Use this skill when the user wants to:
| Feature | SDXL | FLUX |
|---|---|---|
| Speed | ~8s | ~20-80s |
| Negative prompts | YES | No |
| Prompt weights | YES (light) | No |
| Text rendering | Weak | Excellent |
| LoRA ecosystem | Massive | Limited |
Choose SDXL for: Speed, iteration, LoRAs, negative prompt control Choose FLUX for: Quality, text in images, natural language
[Subject], [descriptive details], [style], [quality boosters], [lighting], [composition]
Negative prompt:
[unwanted elements], [quality issues], [style to avoid]
(keyword:1.2) - max 1.48K, highly detailed, sharp focusEssential Boosters:
8K, ultra high resolution, highly detailedsharp focus, intricate detailsprofessional photographycinematic lighting, volumetric lightingPhotography Terms:
DSLR, 50mm lens, f/2.8 aperturesoft natural lighting, golden hour glowshallow depth of field, bokehskin pores visible, fabric textureUniversal (keep short):
low quality, blurry, bad anatomy, extra fingers, deformed hands, watermark
For Realistic Images:
cartoon, illustration, anime, painting, CGI, 3D render, unrealistic proportions, extra fingers, low quality, jpeg artifacts
For Artistic/Illustration:
photorealistic, photograph, 3D render, low quality, blurry, bad composition
./scripts/sdxl-generate.sh "YOUR PROMPT HERE"
./scripts/sdxl-generate.sh "YOUR PROMPT" --n_images 4 --n_rows 2
When user requests an SDXL image:
Understand the request - Subject, style, what to avoid?
Construct prompt pair:
Apply weights sparingly - Only for critical elements, max 1.3-1.4
Generate using the script
Iterate - SDXL's speed allows rapid refinement
Prompt: Portrait of a middle-aged man with salt-and-pepper beard, warm brown
eyes, weathered skin with visible pores, wearing a wool sweater, soft window
light from the left creating gentle shadows, shallow depth of field, intimate
documentary style, 8K, highly detailed
Negative: cartoon, illustration, anime, CGI, plastic skin, extra fingers,
deformed face, low quality, blurry
Prompt: Vast mountain valley at dawn, ancient stone ruins covered in moss in
the foreground, mist rolling through pine forests, snow-capped peaks catching
the first 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
Prompt: Luxury watch on polished black marble surface, dramatic rim lighting,
reflections on metal case, visible texture on leather strap, professional
product photography, commercial quality, sharp focus throughout, studio
lighting with soft gradient background
Negative: low quality, blurry, distorted proportions, floating objects,
cluttered background
Prompt: Young explorer with a glowing lantern standing at the entrance of an
ancient library, towering bookshelves disappearing into shadow, dust motes
floating in light beams, warm amber and cool blue color contrast, detailed
illustration style, storybook aesthetic, intricate linework
Negative: photorealistic, photograph, low quality, simple, flat colors
Prompt: Modern minimalist house with floor-to-ceiling windows, surrounded by
a Japanese garden with carefully raked gravel and sculptural pines, golden
hour light casting long shadows, architectural photography, clean lines,
professional real estate photography style
Negative: cartoon, sketch, low quality, distorted perspective, cluttered
Apply weights sparingly for emphasis:
(glowing lantern:1.3), ancient library, towering bookshelves
Weight guidelines:
1.5: Often ignored or causes artifacts
npx claudepluginhub lando-labs/claude-plugins --plugin model-bridgeEnhances image generation prompts with Subject-Context-Style structure, lighting physics, camera terminology, and character consistency patterns. Useful for creating detailed, physically coherent image prompts.
Generates images from text prompts using fal.ai FLUX.1/2 Pro/Dev/Schnell and SDXL models. Configures image sizes, guidance scale, inference steps, LoRA, seeds, batching via TypeScript/Python APIs.
Provides prompting techniques for AI image generation and editing models on Replicate. Covers natural language prompts, photographic vocabulary, and iterative editing.