Images

A picture that reaches a customer has usually been through several hands: photographed by a supplier, cut out of its background, set on white for a marketplace, warmed up to match the rest of the catalogue, framed for a story, signed with a logo and a price. In Madoo each of those is a node, and a workflow chains them so that the same result comes out every time — for one product, or for the three hundred products of a season. This section explains how the image nodes work and how they combine, from a single resized photo to a campaign generated with AI.

Three ideas run through every page.

Two kinds of step. Most image work — sizing, cropping, colour, filters, compositing, text, layouts, codes — runs on Madoo’s own workers: it is deterministic, gives the same picture for the same input, and consumes no credits. A model is needed only where pixels alone cannot do the job: separating a product from a real background, enlarging with real detail, placing a product in a scene, dressing a model, generating a picture from a prompt. Those AI steps are paid, and they give a proposal — the same prompt does not make the same picture twice.

Local around AI. A good image workflow prepares a picture locally before the AI step (the right size, the right file, a check that it is usable) and finishes it locally afterwards (the exact format, the brand, the text, a final check). The paid step then receives what it needs and nothing is spent on a file that could never work.

Every page has a lab. Each technique comes with a small workflow you can install and run, built on real photographs — objects from the open-access collection of the Metropolitan Museum of Art — so every claim on these pages was measured on a run, not assumed.

Where to go

Deterministic — local, 0 credits

Your task	Read
Bring pictures in and out; formats, upright phone photos, transparency, weight, metadata, colour profiles; measure a picture and hold it back	Image files and formats
Where the subject of a picture is and how big it is: its box, how much of the frame it fills, whether it is cut off, its proportions; which cells of a generated grid hold one	Measuring the subject
The colors of a picture: its palette, dominant and accent colors, the color of a product, a text color that reads on it; whether a picture keeps to the brand colors	The colors of an image
How bright or dark a picture is, whether it is blown out or crushed, whether a packshot sits on white, which area is darker, how bright the subject is apart from its backdrop	Measuring tone and colour
Whether a picture kept its fine detail against a reference — after a generation, an upscale, a compression; where to look closely, cut out at full size for a vision model	Measuring fine detail
Whether a person, an animal or an object is in a picture, how many and where; the mask of the main person to measure or crop it	Finding people and objects
A head shot from any photo — square, 4:5, 9:16 — framed around the face; whether a picture shows a face	Finding and framing faces
This product, this person or this mark found in a scene and cut out, ready to be judged or measured	Finding a subject in a picture
Resize without cutting or stretching what matters; a marketplace square, a story, an email header; crop, trim, frame, turn	Geometry
Correct exposure and colour, give a batch one look — levels, curves, vibrance, colour balance, LUT — and the filters	Colour, tone and filters
A logo in a corner, text in the brand’s fonts, watermarks, shadows, masks, green screens	Compositing and branding
Several pictures into one, the same steps on a whole batch, QR codes and barcodes that scan	Layouts, batches and codes

Generative AI — paid, a proposal

Your task	Read
How an AI image step works: the default model and the others, credits and variants, model settings, presets, effects, prompts, which images reach the model, estimate and charge	How AI image steps work
Generate a picture from a prompt, edit one with several references, fuse products, analyse and classify pictures	Generating and editing with a prompt
Product photography, lifestyle scenes, showrooms, background swap, magic swap, ads	Product imagery
Put a garment on a model, choose poses and settings, and go on to a fashion clip	People and fashion
Show a design on a product as embroidery, foil, engraving or print	Designs on products
Let one node invent the picture — a surprise, one idea per variant	Magic Box
Cut out, upscale, inpaint, remove objects and text, extend or reframe a picture, fit an exact size	AI edits and utilities
Make many pictures look like one shoot and one brand	Consistency and brand
Look at a picture with a model before delivering it or paying for it — focus, exposure, a watermark, the framing — or choose the best of several	Checking pictures
Find out why an image workflow was refused, failed, cost more or delivered something wrong	Troubleshooting

The nodes by family

Family	Nodes	Cost
In, out, measure	Image Input, Image Output, Media from URL, Image Info, Media Technical Info, Measure Subject, Color Palette, Image Statistics, Image Detail, Instance Segment, Face Detect, Quality Gate	0
Geometry	Resize, Resize for Platform, Crop, Auto Crop, Place on Canvas, Extend Canvas, Round Corners, Rotate, Flip, Deskew, Perspective	0
Delivery	Convert Format, Compress, Strip Metadata, Color Profile	0
Tone and colour	Auto Enhance, Brightness & Contrast, Exposure, Levels, Curves, Shadows & Highlights, Hue & Saturation, Vibrance, Temperature & Tint, Colour Balance, Colour LUT, Grayscale	0
Filters	Gaussian, Motion and Radial Blur, Unsharp Mask, Denoise, Vignette, Film Grain, Sepia, Colorize, Gradient Map, Posterize, Threshold, Invert, Oil Paint, Emboss, Edge Detect	0
Compositing	Overlay, Blend, Watermark, Text on Image, Border, Drop Shadow, Apply Mask, Chroma Key, Alpha Set	0
Layouts and codes	Stack, Grid, Collage, QR Code & Barcode	0
Generative AI	Text to Image, Image Transform, Image Fusion, product photography, lifestyle, showroom, background swap, magic swap, ad generator, style transfer, virtual try-on, Magic Box	per image, by model
AI edits and analysis	Remove Background, Upscale, Perfect Size, Outpainting, Reframe, Inpainting, Object Removal, Text Removal, Image Analysis, Image Classifier, Image to Prompt	per image or per call, by model
Judging pictures	Decide with a Model — focus, exposure, watermark, framing, the best of several (Checking pictures)	≈ 0.006 a picture

The exact inputs, outputs and settings of every node are in the catalogue (get_node_type over MCP, or the node types API); the families above are what these pages explain.