bedrock-image-mcp-server

Upscale images to 4K while preserving original details. This tool uses Stability AI's conservative upscaling to increase resolution to 4K while preserving the original image characteristics with minimal alterations. Best for images that already have good quality but need higher resolution. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Key Features - Upscales images to 4K resolution - Preserves original image characteristics - Minimal AI alterations - Input: 64x64 to 9.4 megapixels NOTE ON output_format: png (the default), jpeg and webp are all supported. Prefer jpeg or webp for 1MP inputs — the ~3112x3112 result is about 20MB as a PNG, over Bedrock's 16MB response limit, and fails with "Response payload size ... exceeds the maximum allowed size" even though the input passed validation. PNG succeeds from smaller inputs. ## When to Use - High-quality images that just need more resolution - When you want to preserve the original look - Professional photos that need print resolution - When creative enhancement is not desired ## Example Usage - Image: "product_photo.jpg" Prompt: "professional product photograph, high quality" - Image: "portrait.png" Prompt: "portrait photograph, natural lighting" Returns: McpImageGenerationResponse: A response containing the upscaled image paths.

Generate an image from a text prompt. FASTEST and cheapest option. This tool uses Stable Image Core, Stability AI's fast tier, built on an enhanced SDXL. AWS describes it as delivering exceptional speed and efficiency at consistent quality. It is the quickest and least expensive option here, at lower fidelity than SD3.5 or Ultra. Stability's guidance: ideal for "rapid content generation at scale" and "rapidly iterating on concepts during ideation" — their example is a retailer generating product images for new arrivals. ## Choosing between the text-to-image tools Stability AI's own "ideal for" guidance, with their credit cost per image: - generate_image_ultra (8 credits) — "Photorealistic, Large-Scale Output". Ideal for "ultra-realistic imagery for luxury brands and high-end campaigns", and "professional print media and large format applications". Their example is a luxury brand producing magazine spreads. - generate_image_sd35 (6.5 credits) — "High-Quality, High-Quantity Creative Assets". Ideal for "high-volume outputs like marketing campaigns and digital assets", and "professional use cases at 1 megapixel resolution". Their example is a game team producing environment textures and character concepts. - generate_image_core (3 credits) — "Fast and Affordable". Ideal for "rapid content generation at scale", and "rapidly iterating on concepts during ideation". Their example is a retailer generating product images for new arrivals. So: Ultra for a small number of premium, large-format or print pieces; SD3.5 when you need many good assets; Core when speed and cost dominate. Stability credits only Ultra with typography, so prefer it when the image contains text. ## Requirements and limits - Available in us-west-2 only - Text-to-image only; it cannot transform an existing image. Use transform_image_sd35 for image-to-image work. - output_format must be png or jpeg. webp is rejected by the model. - Returns a single image per call. There is no width/height, cfg_scale or style preset; use aspect_ratio to control the shape. ## Example Usage - Prompt: "three thumbnail concepts for a coffee shop logo, flat vector" Aspect Ratio: 1:1 (Produces a quick draft to iterate on) Returns: McpImageGenerationResponse: A response containing the generated image paths.

Find and replace objects using text prompts. This tool automatically finds specific objects or areas in images using text prompts and replaces them with AI-generated content. No manual masking required. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## CRITICAL: When to Use This Tool **USE THIS TOOL FOR:** - Replacing specific objects (e.g., "replace the chair with a sofa") - Changing ONLY the background (e.g., search_prompt="background", prompt="beach scene") - Swapping elements while keeping the rest unchanged - Product photography variations (e.g., change product color/style) - Targeted edits to specific parts of an image **DO NOT USE THIS TOOL FOR:** - Changing overall image style → Use transform_image_sd35 instead - Just recoloring an object → Use search_and_recolor instead (faster, preserves structure) - Removing objects without replacement → Use remove_object instead ## How It Works 1. search_prompt: Describes what to find (e.g., "background", "chair", "person's shirt") 2. prompt: Describes what to replace it with (e.g., "beach with palm trees", "red leather sofa") 3. The tool automatically detects the object/area and replaces it seamlessly ## Key Features - Automatic object detection via text prompts - No manual masking required - Natural replacement blending - Preserves the rest of the image unchanged ## Example Usage - Image: "room.jpg" Search Prompt: "wooden chair" Prompt: "modern leather armchair" (Finds wooden chair and replaces with leather armchair) - Image: "portrait.jpg" Search Prompt: "background" Prompt: "beautiful beach scene with sand and ocean" (Replaces only the background, keeps subject unchanged) Returns: McpImageGenerationResponse: A response containing the edited image paths.

Remove unwanted objects from images. This tool intelligently removes unwanted objects from images and fills the area with content that blends naturally with surroundings. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Intelligent object removal - Natural background filling - Mask-based control (white=remove, black=preserve) - grow_mask parameter for better edge blending ## When to Use - Removing unwanted objects - Cleaning up photos - Removing distractions - Photo restoration ## Example Usage - Image: "photo.jpg" Mask: "object_mask.png" (Removes object defined by mask) Returns: McpImageGenerationResponse: A response containing the edited image paths.

Apply style from one image to the content of another. This tool applies the style from one image to the content of another, with fine-grained control over composition, style strength, and changes. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Transfers style between images - Fine-grained control over composition and style - Preserves content while applying style - Multiple control parameters for precise results ## Parameters - **init_image**: Path to content/initialization image or base64 data - **style_image**: Path to style reference image or base64 data - **prompt**: Description to guide the transfer - **composition_fidelity**: 0.0-1.0 (default 0.9) - Controls how much of the init_image composition is preserved - Higher = more preservation of original composition - **style_strength**: 0.0-1.0 (default 1.0) - Controls strength of style application - Higher = stronger style application - **change_strength**: 0.0-1.0 (default 0.9) - Controls amount of change allowed - Higher = more changes permitted - **negative_prompt**: Elements to exclude (optional) - **seed**: For reproducible results (0 = random) - **output_format**: jpeg, png, or webp ## When to Use - Artistic style transfer - Photo stylization - Creating variations with different styles - Combining content and style from different sources ## Example Usage - Init Image: "photo.jpg" Style Image: "painting.jpg" Prompt: "transfer impressionist painting style to photograph" Composition Fidelity: 0.9 Style Strength: 1.0 (Applies painting style to photo while preserving composition) Returns: McpImageGenerationResponse: A response containing the generated image paths.

Upscale images to 4K with creative AI enhancement. This tool uses Stability AI's creative upscaling to enhance and upscale images to 4K resolution (20-40x upscale). Best for low-resolution or degraded images that need improvement and enhancement. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Key Features - Upscales images to 4K resolution (20-40x) - AI enhancement adds details and improves quality - Creativity parameter controls enhancement level - Optional style presets for specific aesthetics - Input: 64x64 to 1 megapixel (1024x1024) IMPORTANT: pass output_format="jpeg" (or "webp") for this tool. All three formats are supported and png is the default, but this tool's output is a fixed ~3150x3150 regardless of input size, and a PNG that large exceeds Bedrock's 16MB response limit — so there is no input size at which the default succeeds here. ## Creativity Parameter Guide - 0.1-0.2: Subtle enhancement, mostly preserves original - 0.3: Balanced enhancement (default, recommended) - 0.4-0.5: Strong enhancement, adds significant details ## When to Use - Low-resolution images that need quality improvement - Degraded or compressed images - Images that benefit from AI enhancement - When you want to add artistic details during upscaling ## Style Presets Available presets: 3d-model, analog-film, anime, cinematic, comic-book, digital-art, enhance, fantasy-art, isometric, line-art, low-poly, neon-punk, origami, photographic, pixel-art, tile-texture ## Example Usage - Image: "old_photo.jpg" Prompt: "vintage photograph, restored quality, clear details" Creativity: 0.3 Style: "photographic" Returns: McpImageGenerationResponse: A response containing the upscaled image paths.

Automatically remove backgrounds from images. This tool automatically removes backgrounds from images, isolating the main subject with clean edges. Always outputs PNG with transparency. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Automatic background removal - Clean edge detection - Handles complex subjects (hair, transparent objects) - Always outputs PNG with transparency ## When to Use - Product photography - Portrait isolation - Creating transparent assets - Compositing preparation ## Example Usage - Image: "product.jpg" (Automatically removes background, outputs PNG with transparency) Returns: McpImageGenerationResponse: A response containing the image paths with transparent background.

Create an elliptical mask for inpainting or object removal. This tool creates a grayscale mask image where: - White pixels (255) = areas to fill/remove - Black pixels (0) = areas to preserve The mask is an ellipse centered at (center_x, center_y) with the specified radii. Use this mask with inpaint_image or remove_object tools. ## When to Use - Creating masks for circular or oval objects - Removing round objects (faces, balls, wheels, etc.) - More natural-looking masks than rectangles - Masking organic shapes ## Parameters Guide - **width, height**: Match your source image dimensions - **center_x, center_y**: Center point of the ellipse - **radius_x, radius_y**: Half-width and half-height of the ellipse - **feather**: Add soft edges (5-10 for subtle, 20+ for very soft) ## Example Usage - Image: 1024x768 photo with a face at center Mask: center_x=512, center_y=384, radius_x=100, radius_y=120 (Creates oval mask around the face) Returns: McpImageGenerationResponse: A response containing the mask file path.

Generate an image from a text prompt. BALANCED quality and cost. This tool uses Stable Diffusion 3.5 Large, a strong all-rounder and a good general default. AWS positions it for concept art, visual effects and detailed product imagery across media, gaming, advertising and retail. It renders a wide range of styles (3D, photography, painting, line art) and handles long complex prompts up to 10,000 characters. AWS also credits it with strong text quality, though Stability reserves their typography claim for Ultra. Stability's guidance: ideal for "high-volume outputs like marketing campaigns and digital assets" — websites, newsletters, marketing materials, concept art and game assets. ## Choosing between the text-to-image tools Stability AI's own "ideal for" guidance, with their credit cost per image: - generate_image_ultra (8 credits) — "Photorealistic, Large-Scale Output". Ideal for "ultra-realistic imagery for luxury brands and high-end campaigns", and "professional print media and large format applications". Their example is a luxury brand producing magazine spreads. - generate_image_sd35 (6.5 credits) — "High-Quality, High-Quantity Creative Assets". Ideal for "high-volume outputs like marketing campaigns and digital assets", and "professional use cases at 1 megapixel resolution". Their example is a game team producing environment textures and character concepts. - generate_image_core (3 credits) — "Fast and Affordable". Ideal for "rapid content generation at scale", and "rapidly iterating on concepts during ideation". Their example is a retailer generating product images for new arrivals. So: Ultra for a small number of premium, large-format or print pieces; SD3.5 when you need many good assets; Core when speed and cost dominate. Stability credits only Ultra with typography, so prefer it when the image contains text. Reach for generate_image (Nova Canvas) only when you need explicit pixel dimensions, a color palette, Nova style presets, or several images in one request. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Key Features - Supports prompts up to 10,000 characters (vs 1,024 for Nova Canvas) - Better prompt adherence and detail preservation - Aspect ratio selection instead of explicit dimensions - Reproducible results with seed values ## Prompt Best Practices - Be specific and descriptive with your prompts - Use aspect ratios appropriate for your use case (16:9 for landscapes, 9:16 for portraits, etc.) - Leverage negative prompts to exclude unwanted elements - Use seeds for reproducible results - SD3.5 can handle longer, more detailed prompts effectively ## Example Prompts - "A serene Japanese garden at dawn, with a wooden bridge over a koi pond, cherry blossoms in full bloom, soft morning mist, traditional stone lanterns, and Mount Fuji visible in the distance, photorealistic style" - "Portrait of a wise elderly wizard with a long silver beard, wearing deep blue robes embroidered with golden stars, holding an ancient wooden staff, warm candlelight illuminating his face, fantasy art style" ## Aspect Ratios - 16:9 - Widescreen landscape - 1:1 - Square - 21:9 - Ultra-wide - 2:3 - Portrait - 3:2 - Landscape - 4:5 - Portrait - 5:4 - Landscape - 9:16 - Vertical/mobile - 9:21 - Ultra-tall Returns: McpImageGenerationResponse: A response containing the generated image paths.

Convert sketches or line art into detailed images. This tool converts sketches or line art into detailed images while preserving the structure and composition of the original sketch. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Preserves sketch structure and composition - Adds detail and style based on prompt - Control strength parameter for fine-tuning - Supports various artistic styles ## Parameters - **sketch**: Path to sketch/line art image or base64 data - **prompt**: Description of desired output (style, colors, details) - **control_strength**: 0.0-1.0 (default 0.7) - 0.0 = Ignore sketch, generate freely - 0.7 = Balanced adherence (recommended) - 1.0 = Strict adherence to sketch - **negative_prompt**: Elements to exclude (optional) - **seed**: For reproducible results (0 = random) - **output_format**: jpeg, png, or webp ## When to Use - Converting concept sketches to detailed art - Rapid visualization of ideas - Maintaining specific compositions - Iterating on sketch-based designs ## Example Usage - Sketch: "character_sketch.png" Prompt: "detailed fantasy character, vibrant colors, digital art style" Control Strength: 0.7 (Converts sketch to detailed character art) Returns: McpImageGenerationResponse: A response containing the generated image paths.

Extend images beyond their original boundaries. This tool extends images beyond their original boundaries in specified directions, generating new content that matches the original image style. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Extend images in any direction (left, right, up, down) - Generate content matching original style - Control creativity level - Useful for changing aspect ratios or expanding compositions ## When to Use - Expanding image canvas - Changing aspect ratios - Adding more context to scenes - Creating panoramic views ## Example Usage - Image: "landscape.jpg" Prompt: "mountain peaks and sky" Up: 500 (Extends image upward with mountains and sky) Returns: McpImageGenerationResponse: A response containing the outpainted image paths.

Generate images matching a reference style. This tool generates images matching a reference style while following the content description in the prompt. The fidelity parameter controls how closely the output matches the reference style. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Matches reference image style - Applies style to new content - Fidelity control for style strength - Maintains consistent visual aesthetics ## Parameters - **reference_image**: Path to style reference image or base64 data - **prompt**: Description of desired content - **fidelity**: 0.0-1.0 (default 0.5) - 0.0 = Loose style interpretation - 0.5 = Balanced style matching (recommended) - 1.0 = Very close style match - **negative_prompt**: Elements to exclude (optional) - **seed**: For reproducible results (0 = random) - **output_format**: jpeg, png, or webp ## When to Use - Maintaining consistent visual style across images - Applying artistic styles to new content - Brand consistency in generated images - Style exploration and iteration ## Example Usage - Reference Image: "art_style.jpg" Prompt: "mountain landscape at sunset" Fidelity: 0.7 (Generates landscape in the reference art style) Returns: McpImageGenerationResponse: A response containing the generated image paths.

Recolor specific objects using text prompts. This tool automatically finds objects in images and changes their colors while preserving the object's structure, texture, and details. This is a specialized version of inpainting that focuses only on color changes. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## CRITICAL: When to Use This Tool **USE THIS TOOL FOR:** - Changing ONLY the color of specific objects (e.g., "make the car red") - Product photography color variations (e.g., show shirt in different colors) - Recoloring elements while keeping structure intact - Quick color experiments without changing shape/texture **DO NOT USE THIS TOOL FOR:** - Replacing objects entirely → Use search_and_replace instead - Changing object shape or structure → Use search_and_replace instead - Changing overall image style → Use transform_image_sd35 instead ## How It Works 1. select_prompt: Describes what object to recolor (e.g., "car body", "shirt", "wall") 2. prompt: Describes the desired color/style (e.g., "bright red", "navy blue", "golden yellow") 3. The tool automatically segments the object and recolors it while preserving all details ## Key Features - Automatic object detection via text prompts - Color changes while preserving structure and texture - No manual masking required - Maintains image quality and detail - Faster than search_and_replace for color-only changes ## Example Usage - Image: "car.jpg" Select Prompt: "car body" Prompt: "bright red metallic paint" (Changes car body color to red while keeping all details) - Image: "room.jpg" Select Prompt: "wall" Prompt: "soft beige color" (Recolors the wall without changing texture or structure) Returns: McpImageGenerationResponse: A response containing the recolored image paths.

Create a rectangular mask for inpainting or object removal. This tool creates a grayscale mask image where: - White pixels (255) = areas to fill/remove - Black pixels (0) = areas to preserve The mask is a rectangle positioned at (x, y) with the specified dimensions. Use this mask with inpaint_image or remove_object tools. ## When to Use - Creating masks for inpainting specific rectangular regions - Removing rectangular objects or areas - Testing inpaint/remove_object functionality - Quick mask creation without external tools ## Parameters Guide - **width, height**: Match your source image dimensions - **x, y**: Top-left corner of the area to mask - **mask_width, mask_height**: Size of the masked area - **feather**: Add soft edges (5-10 for subtle, 20+ for very soft) ## Example Usage - Image: 1024x768 photo Mask: x=200, y=150, mask_width=300, mask_height=200 (Creates mask for 300x200 area starting at position 200,150) Returns: McpImageGenerationResponse: A response containing the mask file path.

Generate an image from a text prompt. HIGHEST QUALITY option. This tool uses Stable Image Ultra, Stability AI's flagship text-to-image model. It draws on their top models including SD3.5, and AWS describes it as excelling at typography, intricate compositions, dynamic lighting, vibrant colours and artistic cohesion, producing photorealism with exceptional detail. It costs more per image and is slightly slower than the others. Stability's guidance: ideal for "ultra-realistic imagery for luxury brands and high-end campaigns" and "professional print media and large format applications". Also the only one of the three they credit with typography, so prefer it for text in the image. ## Choosing between the text-to-image tools Stability AI's own "ideal for" guidance, with their credit cost per image: - generate_image_ultra (8 credits) — "Photorealistic, Large-Scale Output". Ideal for "ultra-realistic imagery for luxury brands and high-end campaigns", and "professional print media and large format applications". Their example is a luxury brand producing magazine spreads. - generate_image_sd35 (6.5 credits) — "High-Quality, High-Quantity Creative Assets". Ideal for "high-volume outputs like marketing campaigns and digital assets", and "professional use cases at 1 megapixel resolution". Their example is a game team producing environment textures and character concepts. - generate_image_core (3 credits) — "Fast and Affordable". Ideal for "rapid content generation at scale", and "rapidly iterating on concepts during ideation". Their example is a retailer generating product images for new arrivals. So: Ultra for a small number of premium, large-format or print pieces; SD3.5 when you need many good assets; Core when speed and cost dominate. Stability credits only Ultra with typography, so prefer it when the image contains text. ## Requirements and limits - Available in us-west-2 only - Text-to-image only; it cannot transform an existing image. Use transform_image_sd35 for image-to-image work. - output_format must be png or jpeg. webp is rejected by the model. - Returns a single image per call. There is no width/height, cfg_scale or style preset; use aspect_ratio to control the shape. ## Example Usage - Prompt: "a weathered brass compass on a nautical chart, macro photograph, soft window light" Aspect Ratio: 3:2 (Produces a high-detail final asset) Returns: McpImageGenerationResponse: A response containing the generated image paths.

DEPRECATED. Generate an image using Amazon Nova Canvas with text prompt. This tool uses Amazon Nova Canvas to generate images based on a text prompt. The generated image will be saved to a file and the path will be returned. DEPRECATED: AWS marked Nova Canvas as a Legacy model and retires it on 2026-09-30, after which this tool will stop working. AWS also revokes access after 15 days of inactivity. Use generate_image_sd35 (Stable Diffusion 3.5 Large, in us-west-2) instead, which also gives noticeably better prompt adherence and image quality. Only use this tool when you specifically need a Nova feature that SD3.5 lacks: explicit width/height in pixels, quality/cfg_scale tuning, Nova style presets, or several images in one request. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Prompt Best Practices An effective prompt often includes short descriptions of: 1. The subject 2. The environment 3. (optional) The position or pose of the subject 4. (optional) Lighting description 5. (optional) Camera position/framing 6. (optional) The visual style or medium ("photo", "illustration", "painting", etc.) Do not use negation words like "no", "not", "without" in your prompt. Instead, use the negative_prompt parameter to specify what you don't want in the image. You should always include "people, anatomy, hands, low quality, low resolution, low detail" in your negative_prompt ## Example Prompts - "realistic editorial photo of female teacher standing at a blackboard with a warm smile" - "whimsical and ethereal soft-shaded story illustration: A woman in a large hat stands at the ship's railing looking out across the ocean" - "drone view of a dark river winding through a stark Iceland landscape, cinematic quality" Returns: McpImageGenerationResponse: A response containing the generated image paths.

DEPRECATED. Generate an image using Amazon Nova Canvas with color guidance. This tool uses Amazon Nova Canvas to generate images based on a text prompt and color palette. The generated image will be saved to a file and the path will be returned. DEPRECATED: AWS marked Nova Canvas as a Legacy model and retires it on 2026-09-30, after which this tool will stop working. There is no direct replacement for colour-palette guidance; describe the desired colours in the prompt to generate_image_sd35 instead. IMPORTANT FOR Assistant: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Prompt Best Practices An effective prompt often includes short descriptions of: 1. The subject 2. The environment 3. (optional) The position or pose of the subject 4. (optional) Lighting description 5. (optional) Camera position/framing 6. (optional) The visual style or medium ("photo", "illustration", "painting", etc.) Do not use negation words like "no", "not", "without" in your prompt. Instead, use the negative_prompt parameter to specify what you don't want in the image. ## Example Colors - ["#FF5733", "#33FF57", "#3357FF"] - A vibrant color scheme with red, green, and blue - ["#000000", "#FFFFFF"] - A high contrast black and white scheme - ["#FFD700", "#B87333"] - A gold and bronze color scheme Returns: McpImageGenerationResponse: A response containing the generated image paths.

Transform an existing image using Stable Diffusion 3.5 Large with text guidance. This tool uses SD3.5 to transform existing images based on text prompts. The strength parameter controls how much the output differs from the input image. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## CRITICAL: When to Use This Tool vs Others **USE THIS TOOL FOR:** - Changing overall style (e.g., "make it look like a watercolor painting") - Applying artistic effects to the entire image - Changing lighting, mood, or atmosphere across the whole image - Converting to different art styles (photorealistic to cartoon, etc.) **DO NOT USE THIS TOOL FOR:** - Changing ONLY the background → Use search_and_replace instead - Replacing specific objects → Use search_and_replace instead - Changing colors of specific objects → Use search_and_recolor instead - Removing objects → Use remove_object instead - Extending image boundaries → Use outpaint_image instead This tool transforms the ENTIRE image based on your prompt. It cannot selectively edit only parts of the image (like just the background). For selective edits, use the Stability AI editing tools (search_and_replace, search_and_recolor, etc.). ## Key Features - Transform existing images with text guidance - Control transformation intensity with strength parameter - Supports file paths or base64-encoded images - Minimum input image size: 64px per side ## Strength Parameter Guide - 0.0-0.3: Very subtle changes, mostly preserves input - 0.3-0.5: Moderate changes, clear influence from input - 0.5-0.7: Balanced transformation (recommended starting point) - 0.7-0.9: Strong transformation, input provides structure - 0.9-1.0: Dramatic reimagining, minimal input influence ## Prompt Best Practices - Describe the desired transformation clearly - Use negative prompts to exclude unwanted elements - Start with strength 0.7 and adjust based on results - Use consistent seeds to iterate on results ## Example Transformations - Prompt: "Transform into a watercolor painting style, soft pastel colors, artistic brush strokes" Strength: 0.6 Negative: "photorealistic, sharp edges, digital art" - Prompt: "Convert to a dramatic black and white photograph with high contrast" Strength: 0.5 Negative: "color, washed out, low contrast" Returns: McpImageGenerationResponse: A response containing the transformed image paths.

Fast 4x upscaling without creative enhancement. This tool provides quick 4x resolution increase without AI enhancement. Best for images that need simple resolution increase without style changes. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! The workspace_dir parameter should be set to the directory where the user is currently working so that images are saved to a location accessible to the user. ## Key Features - Fast 4x upscaling - No AI enhancement or style changes - Simple and straightforward - Input: 1024 to 1 megapixel (1024x1024) NOTE ON output_format: png (the default), jpeg and webp are all supported. Output scales 4x with the input, so prefer jpeg or webp for larger inputs — a 1 megapixel input returns 4096x4096, about 35MB as a PNG, over Bedrock's 16MB response limit. PNG works from small inputs (256x256 in gives a 1.4MB PNG; 512x512 gives 5.9MB). ## When to Use - When you need quick upscaling without enhancement - When preserving exact original appearance is critical - For technical images, diagrams, or screenshots - When speed is more important than quality enhancement ## Example Usage - Image: "screenshot.png" (No prompt needed for fast upscale) - Image: "diagram.jpg" (Simple 4x resolution increase) Returns: McpImageGenerationResponse: A response containing the upscaled image paths.

Generate images following structural guides like edge maps or depth maps. This tool generates images that follow structural guides while adding detail and style based on the prompt. Maintains specific compositions. IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Follows structural guidance (edges, depth, etc.) - Maintains composition while adding detail - Control strength for fine-tuning adherence - Supports various structural inputs ## Parameters - **structure_image**: Path to structure/edge map or base64 data - **prompt**: Description of desired output (style, colors, details) - **control_strength**: 0.0-1.0 (default 0.7) - 0.0 = Ignore structure, generate freely - 0.7 = Balanced adherence (recommended) - 1.0 = Strict adherence to structure - **negative_prompt**: Elements to exclude (optional) - **seed**: For reproducible results (0 = random) - **output_format**: jpeg, png, or webp ## When to Use - Maintaining specific compositions - Architectural visualization - Scene layout control - Depth-guided generation ## Example Usage - Structure Image: "edge_map.png" Prompt: "modern architecture, glass and steel, sunset lighting" Control Strength: 0.8 (Generates detailed architecture following edge map) Returns: McpImageGenerationResponse: A response containing the generated image paths.

Fill masked regions with AI-generated content (generative fill). This tool fills masked regions of images with AI-generated content that blends naturally with the surrounding image. The mask defines which areas to fill (white) and which to preserve (black). IMPORTANT FOR ASSISTANT: Always send the current workspace directory when calling this tool! ## Key Features - Fill masked areas with AI-generated content - Natural blending with surrounding image - Mask-based control (white=fill, black=preserve) - grow_mask parameter expands mask edges for better blending ## When to Use - Adding objects to images - Replacing specific regions - Filling in missing or damaged areas - Creative image composition ## Example Usage - Image: "photo.jpg" Mask: "mask.png" Prompt: "a red sports car" (Fills masked area with a red sports car) Returns: McpImageGenerationResponse: A response containing the inpainted image paths.

Create a full white mask covering the entire image. This tool creates a completely white mask useful for: - Complete image replacement with inpainting - Testing inpaint functionality - Full-image generative fill ## When to Use - Testing inpaint_image tool - Replacing entire image content while maintaining dimensions - Quick mask creation for full-image operations ## Example Usage - Image: 1024x768 photo Mask: width=1024, height=768 (Creates full white mask for complete replacement) Returns: McpImageGenerationResponse: A response containing the mask file path.