## The short answer
My Signature trains a personal style model on your own images. You upload 15 to 80 pictures that show the look you want, training runs in the background, and when it is ready it appears as an input node on your canvas. Wire it into a render and the render takes on your style: your light, your palette, your way of staging a building.
## Why a studio would want its own style
Every AI image model has a house look. Ask ten studios for a render of the same house and you get ten images that look like cousins. For a studio with a recognisable way of presenting work, that is a problem. The colour grade, the time of day, the planting, the amount of people and haze are part of how your work is read.
A reference image helps for one render. A trained style helps for all of them, so a whole competition board or a year of projects looks like it comes from the same hand.
## Step 1: choose the images
This is the step that decides the result, so take your time.
- **Use between 15 and 80 images.** More is not always better. Thirty good images beat eighty mixed ones.
- **Pick one look, not a portfolio.** If half your images are warm dusk exteriors and half are cold daylight interiors, the model learns a muddy average. Train two styles instead.
- **Use your own work, or work you have the right to use.** Do not train on other people's images or on a client's confidential material.
- **Avoid text, logos and watermarks.** The model learns whatever is repeated, including a corner logo.
- **Mix the subjects, keep the treatment.** Different buildings and angles, the same light and colour.
You can sort a style into a category when you create it: Exterior Lighting, Interior Atmosphere, Material and Texture, Landscape, or Abstract and Diagram. It helps you find it again.
## Step 2: name it and add a trigger word
Give the Signature a name you will recognise. You can also add a **trigger word**, an unusual word that switches the style on when it appears in a prompt. It is optional, but it is useful if you train several styles and want to ask for each by name.
## Step 3: train
Start training and carry on working. Training usually takes 10 to 20 minutes, and the page shows the status: queued, training, ready or failed. If a job fails you can retry it. A **learning strength** indicator shows how much material the model has to work from, which is a quick check on whether the set was big and varied enough.
## Step 4: use it
When the status is **Ready**, add the **My Signature** node from the Inputs menu of the canvas and wire it into a render node. From then on the render is made in your style. Signatures are FLUX style models, so they apply when you render with a FLUX-based model; the model picker in the render node shows which models are available.
Combine it with your geometry. Send a view from [SketchUp](/plugins/sketchup), [Rhino](/plugins/rhino) or [Revit](/plugins/revit), keep the building and the camera as you modelled them, and let the Signature decide the light and the materials.
## No time to train? Use the style library
If you only need a look, not your own, the Signature page also offers a library of ready-made styles. They need no training and no training slot.
## What it will not do
- It will not copy a single image. A style model learns a treatment, not a picture.
- It will not fix a weak set. Blurry, mixed or heavily edited images give a blurry, mixed result.
- It will not replace judgement. Check the first renders against your own work and retrain with a tighter set if it drifts.
## Plans and slots
My Signature is part of the Pro and Studio plans, and extra training slots can be bought on the Signature page. See the [pricing page](/pricing) for what each plan includes.