What Seam does

Seam adds three tools to Photoshop: Create, Edit, and Upscale. Everything it makes arrives as a normal layer in the document you already have open.

01

Create

Write what you want to see, pick a model, and the image is placed into your document. You set the aspect ratio and how many variations to compare. Reference images are optional. Add some when the result should follow an existing style.

02

Edit

Select part of your image with any Photoshop tool and describe the change: remove an object, add one, swap the background, relight the scene. The result comes back masked to your selection, so everything outside it stays untouched.

03

Upscale

Make an image larger than it started. Specialized models rebuild fine detail at the new size instead of stretching pixels. Useful for print, tight crops, or old material.

01

Preview before commit

Some models render a small draft first, for a fraction of the credits. The full-resolution render runs only once the direction is right.

02

Prompt enhance

One click expands a short prompt into the detailed description models respond best to. The rewrite lands in the prompt field, where you can still edit it.

03

Auto Mask (beta)

After generating, Seam can find the main subject of your prompt and attach an object mask to the new layer, ready for cutouts. If detection fails, you still get your image, just without the mask.

04

Layers that remember

Every generated layer stores the model, prompt, seed, and settings that made it. Select the layer weeks later and load that exact setup back into the plugin.

05

Prompt and style libraries

Keep prompts you reuse as presets. Style libraries hold shared reference images in the cloud, so a team generates against the same visual line.

06

Color profiles

Results are converted into the color profile of your document, whether that is sRGB, Adobe RGB, ProPhoto, or Display P3. Nothing washes out on the way in.

Seam runs no image model of its own. It connects Photoshop to Gemini, GPT Image, FLUX, Reve, and specialist upscalers: one dropdown, one account, one pool of credits. Which model fits depends on the task.

Compare all models