How-to guides

How to add text and layers to an image

Open Image Studio on a generated picture, type editable text straight onto the artwork, style it, and save it as a branch. Layer saves are instant and use no image credits.

Generate an image, then open GPTree Image Studio on it to add words or a logo. The text you add is real editable type, not letters painted by the model, so it is spelled correctly and you can change it afterwards without regenerating anything.

Add a headline

  1. Generate an image, or open one already in the thread.
  2. Open Image Studio from the image.
  3. Add a text layer. It opens with a caret already on the picture, so you can type your copy straight onto the artwork.
  4. Shape it with the controls in the panel, or work directly on the canvas.
  5. Save. This creates a child branch with your layers composed onto the picture.

Layer-only saves are instant and use no image credits, so trying a fifth headline costs nothing but your time.

Work directly on the picture

You do not have to type in the side panel and look across at the image to see what happened. Double-click any text on the canvas to put a caret in it and edit the words in place.

A selected layer carries grips you can drag:

  • Corner grips make the type bigger or smaller.
  • Side grips change how wide the copy runs, so it rewraps without changing its size. Use these when a headline needs to break over two lines.
  • The grip above the layer rotates it. Hold Shift while you drag to turn in fifteen degree steps.

As you drag a layer, guides appear when its edge or centre lines up with the edge of the picture, the middle, the thirds, or another layer, and the layer settles onto that line. Hold Alt while dragging to place it freely instead.

The arrow keys nudge a selected layer one pixel, or ten with Shift held. Cmd+D on a Mac, or Ctrl+D elsewhere, duplicates a layer so a second headline starts from the first.

Cmd+Z undoes anything you just did, including typing, moving, restyling, adding or deleting a layer, and painting a selection. Shift+Cmd+Z redoes it.

Shape the type

Text layers use Manrope, from Regular through Extra bold, or IBM Plex Mono in Medium. The weights offered are the ones that render in the saved picture, so what you preview is what you get.

Beyond font, weight, size and colour you can set italic, capitalisation, horizontal and vertical alignment, line height, letter spacing, padding, outline width and colour, a drop shadow with its own blur, strength, offset and colour, and a solid background behind the copy. Opacity and rotation apply to the whole layer.

Box controls how the text frame behaves. "Hug text" keeps the frame tight around the copy. "Fixed width" holds the width you dragged and rewraps inside it. Dragging a side grip switches a layer to fixed width for you.

Make text readable on a busy picture

White copy over photography is the hardest thing to get right by eye. Select the text layer and press Auto next to Legibility. It sets a soft shadow and a hairline outline sized to your type, picking dark treatment for light text and light treatment for dark text.

Adjust from there if you want more or less. Shadow strength at zero turns the shadow off entirely, and outline at zero removes the outline.

Add a logo or a second image

Add an image layer instead of a text layer, then scale, rotate, and position it. This is the reliable way to get a real logo onto a picture. Describing your logo in a prompt gives the model nothing to reproduce, so it invents something close and wrong.

One thing to know: the layer keeps pointing at the file you added. If that image is later removed, the edit will refuse rather than quietly dropping the layer, and the message will name the problem so you can delete that layer and continue.

Ask the model for a visual change

Use "Describe the change" when the picture itself needs to change, for example a different background or an object removed. That request goes to the image model and costs credits, the same as generating an image.

To aim a change at one area, paint over it with the brush first, adjust the brush size as needed, and use Refine selection to tighten the edge. The mask has its own undo and redo.

Set your expectations for masking correctly. Current image models rebuild the whole frame rather than editing pixels in place, so the mask is a strong steer rather than a hard edge. Usually the change lands where you painted; sometimes something outside it shifts too. If that happens, the branch you started from is untouched, so go back and try a different wording.

When a model change succeeds, GPTree reapplies your editable layers over the new visual base. Your headline survives the background change.

Compare several treatments

Because every save is a branch rather than an overwrite, you can build several versions from one generated picture and keep them all. Make one headline, go back to the original, make a different one, and both stay. The Branch history panel lists them so you can move between them and pick the strongest.

This is usually faster than trying to get one perfect image out of the model, and it leaves you with the alternatives to show someone rather than a single take.

If you hit the layer limit

One edit can carry up to eight layers. Past that the edit stops and tells you how many it needed, rather than silently dropping the extras and handing you an image missing the words you asked for. Delete a layer you no longer need and save again.

If the text still looks wrong

Check that you added a text layer rather than asking the model for the words in the prompt. Anything typed into "Describe the change" is drawn by the image model and will have the usual problems with spelling and letterforms. Text layers are set by GPTree and come out exactly as typed.

Related reading