Skip to main content

Module 6 โ€ข Lesson 1๐Ÿช„ Generative Fill

Welcome to the Advanced tier. Generative Fill lets you add or remove parts of an image by describing them in plain words, and Photoshop paints the result onto a new layer. It feels like magic, but it is a tool with rules, strengths, and blind spots. This lesson teaches you to drive it well.

๐Ÿ“š What You'll Learn

By the end of this lesson, you will be able to:

  • Remove an object by selecting it and generating with an empty prompt
  • Add new content by selecting an area and describing what belongs there
  • Work with generative layers and cycle through variations
  • Write prompts that work, and know Generative Fill's limits

โฑ๏ธ Estimated Time: 50 minutes

๐ŸŽฏ Project: Take one photo and both remove a distraction and add a believable new element, non-destructively.

In This Lesson

โœจ The Goal: Describe It Into Being

Same lake, two edits at once. On the right, an ugly sign clutters the shore and the water is empty. On the left, the sign is gone (removed with an empty prompt) and a sailboat has appeared (added with the prompt "a small sailboat"). Drag the handle:

After Before
Two generative edits: a removal (empty prompt on the sign) and an addition (a described sailboat), each on its own generated layer.
Illustrated stand-in. Real generated results drop into this same slider.

๐Ÿง  Mental Model: Select, Then Describe

Generative Fill always works the same way, and it is worth saying out loud because it makes everything predictable: you select an area, then you describe what should be there. Photoshop looks at the surrounding image for context, generates content to fit, and places it on a brand-new generative layer with its own mask. Your original pixels are never touched.

Two everyday jobs come straight out of that one behavior:

  • Remove: select an object and generate with an empty prompt. Photoshop fills the selection with plausible background, a smarter cousin of Content-Aware Fill from Module 1.
  • Add: select an empty area and describe what to put there. Photoshop generates it in the scene's light and perspective.

๐Ÿง  It is a layer, not a commitment

Because the result lands on a generative layer, everything you know applies: hide it, mask it, lower its opacity, delete it, or stack another generation on top. Generative Fill is not a one-way button. It is one more non-destructive layer, the principle from Module 0 all over again.

flowchart LR S["Select an area<br/>(the where)"] --> P["Type a prompt<br/>(the what)"] P --> G["Generate"] G --> V["3 variations<br/>on a generative layer"] V --> C["Keep one, regenerate,<br/>or mask it"]
Figure 1: The universal loop. Select where, describe what, generate, then choose from the variations. Empty prompt = remove.

๐ŸŽ›๏ธ The Generative Task Bar

When you make a selection, a floating Contextual Task Bar appears with a prompt field and a Generate button. Here it is as a diagram:

Ps   Contextual Task Bar
a small sailboat Generate โ†‘ selected area (the "where")
Figure 2: Select the area, type what belongs there (or leave it empty to remove), and click Generate. Leaving the prompt empty tells Photoshop to fill with background.

๐Ÿ› ๏ธ Guided Build: Remove & Add

Open a photo with a clear distraction and some open space. Save as generative-working.psd. (Generative Fill needs a recent Photoshop, an internet connection, and an Adobe account; it spends generative credits.)

Step 1: Remove a distraction ยท 8 min

  1. Select the object with any selection tool. A loose selection with a little margin works best.
  2. In the task bar, leave the prompt empty and click Generate.
  3. Photoshop fills the selection with background on a new generative layer. Cycle the variations and pick the cleanest.

Step 2: Add a new element ยท 8 min

  1. Select an empty area where something could believably go (open water, sky, a table top).
  2. Type a short, concrete prompt: a small sailboat, a coffee cup, storm clouds. Click Generate.
  3. Review the three variations. Keep one, or click Generate again for three more.

โœ… Match the selection to the subject

Want a tall element? Make a tall selection. Generative Fill fills the shape you give it, so the selection's size and proportion strongly steer the result. Selecting the right space is half the prompt.

Step 3: Blend and refine ยท 6 min

  1. Because it is a normal layer with a mask, refine it with everything you know: mask a rough edge (Module 1), match its light and color with clipped adjustments (Module 4), heal a small artifact (Module 3).
  2. Generative results are a strong starting point, not always a finished one. Your craft finishes the job.

โœ… Project Completion Checklist

  • โ˜ A distraction is removed via an empty-prompt generation
  • โ˜ A new element is added via a described generation
  • โ˜ Each result sits on its own generative layer (original untouched)
  • โ˜ I cycled variations and picked the best
  • โ˜ I refined the result with masks or adjustments as needed

โš–๏ธ Prompts, Limits & Ethics

Prompts that work

  • Be concrete and short. "a wooden park bench" beats "something nice to sit on."
  • Describe the thing, not the style. The scene's existing light usually wins, so name the object and let Photoshop match the lighting.
  • Iterate. If three variations miss, tweak one or two words and generate again rather than rewriting from scratch.

โš ๏ธ Know the limits

  • Fine detail (hands, text, faces at small size) can come out wrong. Zoom in and check.
  • Very large or very specific generations are less reliable than small, plausible ones.
  • It needs internet and credits, and results vary run to run.

๐Ÿ’ก A note on honesty

Generative edits can change what a photo says, not just how it looks. For journalism, evidence, or documentary work, adding or removing content crosses a line. Adobe embeds Content Credentials to flag AI edits for exactly this reason. Use the power thoughtfully, and be transparent when it matters.

๐Ÿง— Now You: Solo Variation

๐ŸŒŸ Your challenge

  1. Remove three different distractions from one busy photo, one selection at a time.
  2. Add an element and then change your mind: regenerate with a different prompt on the same selection. Feel how cheap iteration is.
  3. Try a deliberately hard prompt (a person's hands, readable text) and observe where Generative Fill struggles. Knowing its weak spots makes you a better driver.

Going further: generate an element, then treat it as a real subject: mask it, grade it, and add a shadow with your Module 4 skills. Generative plus craft beats generative alone every time.

๐Ÿณ Recipe Card: Generative Fill

Add or remove by describing

  1. Select the area (loose, with margin)
  2. Remove: empty prompt โ†’ Generate
  3. Add: short concrete prompt โ†’ Generate
  4. Cycle the three variations; regenerate if needed
  5. Refine on its generative layer (mask, match, heal)

Mantra: select the where, describe the what, then finish with craft.

๐Ÿ““ Learning Journal

Add to your journal after this lesson:

  • Key concepts you learned
  • Techniques that clicked for you
  • Questions or confusion points to revisit
  • Ideas you want to try
  • Your progress and feelings about learning this

โœ๏ธ This lesson's prompt: Generative Fill is powerful and a little unsettling. Where is your personal line between "cleaning up a photo" and "changing the truth of it"? Write it down; it is worth knowing before a client asks.

๐Ÿ“ Lesson Summary

๐ŸŽ“ Key Takeaways

  • Generative Fill = select an area, then describe what belongs there.
  • Empty prompt removes; a described prompt adds.
  • Results land on a generative layer, fully non-destructive and refinable.
  • Prompts should be short and concrete; know the limits and edit ethically.

๐ŸŽ‰ What You've Accomplished

You can now add and remove content by describing it, and just as importantly, you treat the result as a layer to refine rather than a magic final answer. That combination, generative power plus real editing craft, is exactly what separates a thoughtful editor from someone mashing a button.

โ“ Common Questions at This Stage

Generative Fill is grayed out. Why?

You likely need a current version of Photoshop, an internet connection, and to be signed in to your Adobe account. The feature runs in the cloud and uses generative credits.

My generation looks blurry or low-detail.

Very large selections spread the generated detail thin. Generate in smaller pieces, or generate then sharpen and refine. Also try regenerating; quality varies between runs.

Can I use Generative Fill results commercially?

Adobe designed Firefly (which powers it) to be commercially safe, but always check the current terms for your plan and region, and disclose AI edits where honesty matters.

๐Ÿ”ญ Looking Ahead

Generative Fill works inside the frame. Lesson 6.2 works outside it: Generative Expand extends the canvas itself, so you can uncrop a photo, fix a tight composition, or change aspect ratio and let Photoshop invent the new space.

โœ… Before the Next Lesson

  • Remove five distractions and add three elements across a few photos.
  • Keep a short list of prompts that worked well for you.
  • Write your Learning Journal entry.

๐Ÿ“š Additional Resources

๐ŸŒŸ Encouragement for the Journey

You just used one of the most powerful features in Photoshop's history, and you used it like an editor, not a slot machine. That mindset, power in service of craft, is exactly what will keep your work valuable as the tools keep getting smarter.